{"items":[{"id":"wangzhang/gemma-4-31B-it-abliterated","name":"wangzhang/gemma-4-31B-it-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69d21deabbf83b1a4bacab73","favorites":39,"downloads":287925,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The wangzhang/gemma-4-31B-it-abliterated model is a 31 billion parameter instruction-tuned variant of Google's Gemma 4 architecture, developed by Wangzhang Wu. This version has been 'abliterated' using direct weight editing techniques to significantly reduce refusal rates, achieving 7/100 refusals compared to 99/100 for the original model on a private evaluation dataset. It is optimized for use cases requiring less restrictive content generation, making it suitable for research and applications where the original model's safety guardrails are deemed too conservative.","is_gated":false,"is_private":false,"is_conversational":false,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":31000000000,"context_length":32768,"created_at":"2026-07-06T08:34:37.770Z","updated_at":"2026-10-15T00:56:44.041Z","hf_created_at":"2026-04-05T08:31:38.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":5322},"classified_tags":{"languages":[],"tasks":[],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["gemma4","direct-weight-editing","abliterix","llm-judge"],"base_models":["google/gemma-4-31b-it","finetune:google/gemma-4-31b-it"]},"model_class":{"id":"gemma4-31b","created_at":"2026-04-08T07:52:17.866Z","updated_at":"2026-06-12T21:16:31.087Z","concurrency_cost":2,"size":31,"parameter_size":31000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma4","family_name":"Gemma 4"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.5-27B-abliterated","name":"huihui-ai/Huihui-Qwen3.5-27B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69a136ac3ccaf5a7691257c4","favorites":157,"downloads":123733,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":null,"acc_tags":[],"description":"Huihui-Qwen3.5-27B-abliterated is a 27 billion parameter language model developed by huihui-ai, based on the Qwen/Qwen3.5-27B architecture with a 32768 token context length. This model has undergone an \"abliteration\" process to significantly reduce its safety filtering, making it an uncensored version of the original. It is primarily intended for research and experimental use in controlled environments where the removal of refusal behaviors is desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":27000000000,"context_length":32768,"created_at":"2026-05-14T20:10:35.513Z","updated_at":"2026-10-14T22:27:26.863Z","hf_created_at":"2026-02-27T06:16:12.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":8609},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3.5-27b","finetune:qwen/qwen3.5-27b"]},"model_class":{"id":"qwen3.5-27b","created_at":"2026-03-08T12:08:17.251Z","updated_at":"2026-03-08T12:08:17.251Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen35","family_name":"Qwen 3.5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.8-27B-abliterated","name":"huihui-ai/Huihui-Qwen3.8-27B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a81734082d78ce589a6b0bb","favorites":569,"downloads":41540,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3.8-27B-abliterated is a 27 billion parameter language model based on the Qwen3.8 architecture, developed by huihui-ai. This model has undergone 'abliteration' to remove refusal behaviors, making it an uncensored version of the original Qwen3.8-27B. It is designed for use cases requiring less restrictive content filtering, while retaining the original model's performance in other areas.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":true,"isHotRecent":true,"loadingStage":null,"sortRank":100,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-16T23:00:30.252Z","updated_at":"2026-10-14T23:13:17.381Z","hf_created_at":"2026-08-16T08:22:24.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":373},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5","qwen3"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["huihui"],"base_models":["qwen/qwen3.8-27b","finetune:qwen/qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"BennyDaBall/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","name":"BennyDaBall/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","owned_by":"Feather","root_model_id":null,"huggingface_id":"6979690431f39715e0819a85","favorites":125,"downloads":35413,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"BennyDaBall/Qwen3-4b-Z-Image-Turbo-AbliteratedV1 is an abliterated version of the Z-Image-Turbo text encoder, specifically modified to reduce generation refusals. Developed by BennyDaBall, this model targets both image generation and general refusals, achieving a low refusal rate of 4/100 in torture tests with a minimal KL Divergence of 0.0004. It is designed to enable more permissive content generation, making it suitable for applications requiring fewer content restrictions.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":4000000000,"context_length":32768,"created_at":"2026-01-29T09:06:39.195Z","updated_at":"2026-10-14T21:42:38.862Z","hf_created_at":"2026-01-28T01:40:20.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":8443},"classified_tags":{"languages":[],"tasks":["conversational"],"architectures":["qwen3","qwen"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic","z-image","censorship-removal"],"base_models":["tongyi-mai/z-image-turbo","quantized:tongyi-mai/z-image-turbo"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3-8B-abliterated-v2","name":"huihui-ai/Huihui-Qwen3-8B-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"6852da2b425c3b44e96fa432","favorites":66,"downloads":23941,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3-8B-abliterated-v2 is an 8 billion parameter Qwen3-based causal language model with a 32768 token context length. Developed by huihui-ai, this model has undergone 'abliteration' to significantly reduce safety filtering and refusals, making it an uncensored version of the original Qwen3-8B. It is primarily designed for research and experimental use cases where unfiltered content generation is desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":8000000000,"context_length":32768,"created_at":"2025-06-28T03:47:07.583Z","updated_at":"2026-10-15T01:22:16.929Z","hf_created_at":"2025-06-18T15:24:27.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7717},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-8b","finetune:qwen/qwen3-8b"]},"model_class":{"id":"qwen3-8b","created_at":"2025-05-12T23:45:01.939Z","updated_at":"2026-01-27T08:03:30.655Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Llama-3.3-70B-Instruct-abliterated","name":"huihui-ai/Llama-3.3-70B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"67590c14e163534a8b199b8b","favorites":90,"downloads":18795,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3.3","categories":null,"tags":null,"acc_tags":[],"description":"huihui-ai/Llama-3.3-70B-Instruct-abliterated is a 70 billion parameter instruction-tuned causal language model derived from Meta's Llama-3.3-70B-Instruct, featuring a 32768 token context length. This model has been specifically modified using an \"abliteration\" technique to remove refusal behaviors, aiming to provide uncensored responses. It is primarily designed for use cases requiring direct answers without built-in content moderation or refusal mechanisms.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":true,"isHotRecent":false,"loadingStage":null,"sortRank":100,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":70000000000,"context_length":32768,"created_at":"2024-12-16T00:00:26.500Z","updated_at":"2026-10-15T00:57:23.983Z","hf_created_at":"2024-12-11T03:50:44.000Z","trending_rank":{"fetched_at":"2026-10-10T00:17:00.027Z","position":4416},"classified_tags":{"languages":["en","fr","it","pt","hi","es","th","de"],"tasks":["text-generation","conversational"],"architectures":["llama","llama-3"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["meta-llama/llama-3.3-70b-instruct","finetune:meta-llama/llama-3.3-70b-instruct"]},"model_class":{"id":"llama33-70b","created_at":"2025-10-15T03:53:00.000Z","updated_at":"2026-01-27T07:36:45.252Z","concurrency_cost":4,"size":70,"parameter_size":70000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama33","family_name":"Llama 3.3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated","name":"huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"68f0acd4c8236d90909e6ec3","favorites":112,"downloads":15781,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated model is a 4 billion parameter vision-language instruction model, based on the Qwen3-VL architecture. This version has undergone 'abliteration' to remove refusal behaviors in its text generation capabilities, making it an uncensored variant of the original Qwen/Qwen3-VL-4B-Instruct. It retains its vision capabilities while offering unrestricted text outputs, suitable for research and experimental use where content filtering is not desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":4000000000,"context_length":32768,"created_at":"2026-05-07T20:04:42.880Z","updated_at":"2026-10-14T22:21:40.963Z","hf_created_at":"2025-10-16T08:29:08.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":5087},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_vl"],"base_models":["qwen/qwen3-vl-4b-instruct","finetune:qwen/qwen3-vl-4b-instruct"]},"model_class":{"id":"qwen3vl-4b","created_at":"2026-05-07T20:01:08.226Z","updated_at":"2026-05-07T20:01:08.226Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"mlabonne/NeuralDaredevil-8B-abliterated","name":"mlabonne/NeuralDaredevil-8B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6654e003965ea394ee6aa251","favorites":297,"downloads":14960,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3","categories":null,"tags":["niche"],"acc_tags":[],"description":"mlabonne/NeuralDaredevil-8B-abliterated is an 8 billion parameter DPO fine-tune of mlabonne/Daredevil-8B-abliterated, designed to recover performance loss from the abliteration process. This model excels as an uncensored language model, making it suitable for applications that do not require alignment, such as role-playing. It demonstrates strong performance on the Open LLM Leaderboard, ranking as the best-performing uncensored 8B model by MMLU score.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":8000000000,"context_length":8192,"created_at":"2024-07-04T08:04:58.801Z","updated_at":"2026-10-15T00:52:45.724Z","hf_created_at":"2024-05-27T19:33:23.000Z","trending_rank":{"fetched_at":"2026-10-07T19:17:00.030Z","position":3932},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":["dpo"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"datasets":["mlabonne/orpo-dpo-mix-40k"],"base_models":["mlabonne/daredevil-8b-abliterated","finetune:mlabonne/daredevil-8b-abliterated"]},"model_class":{"id":"llama3-8b","created_at":"2025-12-02T14:43:00.000Z","updated_at":"2026-03-06T06:13:02.721Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama3","family_name":"Llama 3"},"supports_tool_calling":true},{"id":"prithivMLmods/Gliese-Qwen3.5-9B-Abliterated-Caption","name":"prithivMLmods/Gliese-Qwen3.5-9B-Abliterated-Caption","owned_by":"Feather","root_model_id":null,"huggingface_id":"69afd31197e57fcb354ccda2","favorites":62,"downloads":10619,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The prithivMLmods/Gliese-Qwen3.5-9B-Abliterated-Caption is a 9 billion parameter vision-language model built on Qwen/Qwen3.5-9B, specifically designed for generalized and unfiltered image captioning. It utilizes advanced refusal direction analysis and abliterated training to minimize internal refusal behaviors, maximizing descriptive capability and visual understanding. This model excels at generating highly detailed, context-aware captions and rich visual descriptions from images. It is optimized for tasks requiring comprehensive visual analysis and long-form caption generation.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":9000000000,"context_length":32768,"created_at":"2026-06-15T22:16:48.616Z","updated_at":"2026-10-15T01:16:38.264Z","hf_created_at":"2026-03-10T08:15:13.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":8696},"classified_tags":{"languages":["en","zh"],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["uncensored","abliterated"],"capabilities":["reasoning","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["unredacted","refusal-ablated","bf16","max","alignment-modified"],"base_models":["qwen/qwen3.5-9b","quantized:qwen/qwen3.5-9b"]},"model_class":{"id":"qwen3_5-9b","created_at":"2026-06-12T10:59:16.626Z","updated_at":"2026-06-12T10:59:16.626Z","concurrency_cost":1,"size":9,"parameter_size":9000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"KridgeDookie/KAT-Coder-V2.5-Dev-35B-A3B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS","name":"KridgeDookie/KAT-Coder-V2.5-Dev-35B-A3B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a75e4e29ac03965e03d4940","favorites":6,"downloads":9516,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"KridgeDookie/KAT-Coder-V2.5-Dev-35B-A3B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS is a 35.1 billion parameter Mixture of Experts (MoE) model with approximately 3 billion active parameters, derived from Kwaipilot/KAT-Coder-V2.5-Dev. This text-only model is specifically modified to significantly reduce refusal behavior while preserving its strong coding and agentic coding workflow capabilities. It offers a 32768 token context length and is optimized for developers seeking a highly compliant coding assistant.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":35100000000,"context_length":32768,"created_at":"2026-08-09T12:43:31.438Z","updated_at":"2026-10-15T00:16:46.558Z","hf_created_at":"2026-08-07T14:00:02.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":10189},"classified_tags":{"languages":["en","zh"],"tasks":["image-text-to-text","conversational","text-generation"],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["moe","code","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_5_moe","kat-coder","kat-coder-v2.5","qwen3.6","qwen3.6-35b-a3b","35b","a3b","agentic-coding","coding-agent","refusal-reduction","philadelphia-class"],"base_models":["kwaipilot/kat-coder-v2.5-dev","quantized:kwaipilot/kat-coder-v2.5-dev"]},"model_class":{"id":"qwen3_5moe-35b","created_at":"2026-06-13T11:08:45.209Z","updated_at":"2026-06-13T11:08:45.209Z","concurrency_cost":2,"size":35.1,"parameter_size":35100000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3-5-moe","family_name":"Qwen3 5 Moe"},"supports_tool_calling":true},{"id":"failspy/Meta-Llama-3-8B-Instruct-abliterated-v3","name":"failspy/Meta-Llama-3-8B-Instruct-abliterated-v3","owned_by":"Feather","root_model_id":null,"huggingface_id":"664ac5c0c88927e6ddc48323","favorites":66,"downloads":9220,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3","categories":null,"tags":["niche"],"acc_tags":[],"description":"failspy/Meta-Llama-3-8B-Instruct-abliterated-v3 is an 8 billion parameter instruction-tuned Llama 3 model developed by failspy. It features orthogonally ablated bfloat16 safetensor weights, specifically manipulated to inhibit refusal behaviors without altering other core functionalities. This model is optimized for use cases requiring an uncensored response style while retaining the original Llama 3's knowledge and training.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":8000000000,"context_length":8192,"created_at":"2024-06-13T04:51:01.594Z","updated_at":"2026-10-15T00:47:58.739Z","hf_created_at":"2024-05-20T03:38:40.000Z","trending_rank":{"fetched_at":"2026-10-13T23:17:00.029Z","position":6934},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"llama3-8b","created_at":"2025-12-02T14:43:00.000Z","updated_at":"2026-03-06T06:13:02.721Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama3","family_name":"Llama 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.5-4B-abliterated","name":"huihui-ai/Huihui-Qwen3.5-4B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69ac95b1cb2cff6bffb926fe","favorites":43,"downloads":8772,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3.5-4B-abliterated by huihui-ai is a 4.5 billion parameter language model, fine-tuned from Qwen/Qwen3.5-4B using TRL. This model is specifically designed with significantly reduced safety filtering, making it an uncensored version for research and experimental use. It processes text-based inputs and is intended for applications where content generation without typical safety constraints is desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":4500000000,"context_length":32768,"created_at":"2026-06-15T22:02:56.190Z","updated_at":"2026-10-14T22:58:24.069Z","hf_created_at":"2026-03-07T21:16:33.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":5256},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3.5-4b","quantized:qwen/qwen3.5-4b"]},"model_class":{"id":"qwen3_5-4b","created_at":"2026-06-12T12:56:36.307Z","updated_at":"2026-06-12T12:56:36.307Z","concurrency_cost":1,"size":4.5,"parameter_size":4500000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"mlabonne/Daredevil-8B-abliterated","name":"mlabonne/Daredevil-8B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"665348155ff9a88671a3d02c","favorites":63,"downloads":8670,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"other","categories":null,"tags":["niche"],"acc_tags":[],"description":"mlabonne/Daredevil-8B-abliterated is an 8 billion parameter uncensored language model based on mlabonne/Daredevil-8B, developed using failspy's notebook. It leverages a technique to mediate refusal in LLMs, making it suitable for applications not requiring alignment. This model is optimized for role-playing and similar use cases, demonstrating strong performance on the Open LLM Leaderboard for its size class.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":8000000000,"context_length":8192,"created_at":"2024-07-04T08:05:53.840Z","updated_at":"2026-10-15T00:52:46.609Z","hf_created_at":"2024-05-26T14:32:53.000Z","trending_rank":{"fetched_at":"2026-09-20T12:17:00.020Z","position":3673},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["mlabonne/daredevil-8b","finetune:mlabonne/daredevil-8b"]},"model_class":{"id":"llama3-8b","created_at":"2025-12-02T14:43:00.000Z","updated_at":"2026-03-06T06:13:02.721Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama3","family_name":"Llama 3"},"supports_tool_calling":true},{"id":"failspy/llama-3-70B-Instruct-abliterated","name":"failspy/llama-3-70B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"66397aba7256327c68940bca","favorites":152,"downloads":8655,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3","categories":null,"tags":["niche"],"acc_tags":[],"description":"The failspy/llama-3-70B-Instruct-abliterated model is a 70 billion parameter instruction-tuned language model, derived from Meta's Llama-3-70B-Instruct. This model has undergone specific weight manipulation to orthogonalize the refusal direction, aiming to inhibit the model's tendency to express refusal or lecture on ethics. It maintains the original Llama-3-70B-Instruct tuning in all other aspects, offering an 8192-token context length. Its primary differentiator is the experimental reduction of refusal behaviors, making it suitable for use cases where direct responses are preferred over ethical caveats.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":70000000000,"context_length":8192,"created_at":"2024-06-21T20:24:06.404Z","updated_at":"2026-10-15T00:49:01.327Z","hf_created_at":"2024-05-07T00:50:02.000Z","trending_rank":{"fetched_at":"2026-09-25T17:17:00.022Z","position":3850},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"llama3-70b","created_at":"2024-06-21T16:42:58.361Z","updated_at":"2026-01-27T07:58:09.461Z","concurrency_cost":4,"size":70,"parameter_size":70000000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama3","family_name":"Llama 3"},"supports_tool_calling":true},{"id":"failspy/Meta-Llama-3-70B-Instruct-abliterated-v3.5","name":"failspy/Meta-Llama-3-70B-Instruct-abliterated-v3.5","owned_by":"Feather","root_model_id":null,"huggingface_id":"665619a02fd2e8a25b161b82","favorites":55,"downloads":8612,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3","categories":null,"tags":["niche"],"acc_tags":[],"description":"failspy/Meta-Llama-3-70B-Instruct-abliterated-v3.5 is a 70 billion parameter instruction-tuned causal language model derived from Meta-Llama-3-70B-Instruct. This model utilizes an orthogonalization methodology to specifically inhibit refusal behaviors, making it \"uncensored\" without altering other core functionalities. It maintains the original model's knowledge and training while surgically removing the tendency to refuse user requests, offering a distinct approach compared to traditional fine-tuning.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":70000000000,"context_length":8192,"created_at":"2024-06-20T19:15:06.037Z","updated_at":"2026-10-15T00:48:59.540Z","hf_created_at":"2024-05-28T17:51:28.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"llama3-70b","created_at":"2024-06-21T16:42:58.361Z","updated_at":"2026-01-27T07:58:09.461Z","concurrency_cost":4,"size":70,"parameter_size":70000000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama3","family_name":"Llama 3"},"supports_tool_calling":true},{"id":"failspy/Smaug-Llama-3-70B-Instruct-abliterated-v3","name":"failspy/Smaug-Llama-3-70B-Instruct-abliterated-v3","owned_by":"Feather","root_model_id":null,"huggingface_id":"664ab3f2b9aa0e198384cbf7","favorites":16,"downloads":8566,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3","categories":null,"tags":["niche"],"acc_tags":[],"description":"failspy/Smaug-Llama-3-70B-Instruct-abliterated-v3 is a 70 billion parameter instruction-tuned model based on abacusai/Smaug-Llama-3-70B-Instruct, featuring an 8192-token context length. This model has undergone a unique \"abliteration\" process using orthogonalization to specifically inhibit refusal behaviors, making it effectively uncensored without altering other core functionalities. It is designed for use cases where direct, unfiltered responses are preferred, maintaining the original model's knowledge and training while removing specific undesirable tendencies.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":70000000000,"context_length":8192,"created_at":"2024-07-04T08:03:50.905Z","updated_at":"2026-10-15T00:52:31.476Z","hf_created_at":"2024-05-20T02:22:42.000Z","trending_rank":{"fetched_at":"2026-10-13T23:17:00.029Z","position":6933},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"llama3-70b","created_at":"2024-06-21T16:42:58.361Z","updated_at":"2026-01-27T07:58:09.461Z","concurrency_cost":4,"size":70,"parameter_size":70000000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama3","family_name":"Llama 3"},"supports_tool_calling":true},{"id":"AEON-7/Gemma-4-12B-it-AEON-Abliterated-K4-BF16","name":"AEON-7/Gemma-4-12B-it-AEON-Abliterated-K4-BF16","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a2130880999b80e6a7c3a71","favorites":32,"downloads":6950,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"gemma","categories":null,"tags":["niche"],"acc_tags":[],"description":"AEON-7/Gemma-4-12B-it-AEON-Abliterated-K4-BF16 is a 12 billion parameter instruction-tuned causal language model from AEON-7, based on Google's Gemma-4 architecture. This BF16 variant is a refusal-removed, capability-preserving 'abliteration' of the base model, utilizing a K=4 multi-direction biprojection. It maintains the base model's performance across reasoning, coding, and instruction-following tasks while removing refusal behaviors, making it suitable for fine-tuning or deployment on non-Blackwell hardware where bit-exact weights are preferred.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":12000000000,"context_length":32768,"created_at":"2026-07-12T10:12:38.908Z","updated_at":"2026-10-14T22:57:34.107Z","hf_created_at":"2026-06-04T08:00:08.000Z","trending_rank":{"fetched_at":"2026-10-08T19:17:00.030Z","position":6538},"classified_tags":{"languages":["en"],"tasks":["image-text-to-text","conversational","text-generation"],"architectures":["gemma"],"training":["abliterated","uncensored"],"capabilities":["agent","chat","code","function-calling","instruct","long context","multimodal","reasoning","thinking","vision","vision-language","tool-use"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["gemma4_unified","aarch64","aeon","aeon-7","arm64","bf16","biprojection","blackwell","capability-preserving","chunked-prefill","dense","dgx-spark","gb10","gemma-4","gemma-4-12b","gemma4","gpu","grace-blackwell","heretic","k4-biprojection","low-drift","multi-direction-biprojection","multimodal-capable","openai-api","openai-compatible","prefix-caching","production-ready","refusal-removed","sm_121a","trevor-js"],"base_models":["google/gemma-4-12b-it","finetune:google/gemma-4-12b-it"]},"model_class":{"id":"gemma4unified-12b","created_at":"2026-07-10T12:42:57.508Z","updated_at":"2026-07-10T12:42:57.508Z","concurrency_cost":1,"size":12,"parameter_size":12000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma4-unified","family_name":"Gemma4 Unified"},"supports_tool_calling":true},{"id":"mlabonne/gemma-3-27b-it-abliterated","name":"mlabonne/gemma-3-27b-it-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"67d7262eaf109e28c7ddcc49","favorites":344,"downloads":6921,"avg_rating":5,"total_reviews":1,"status":"active","health":null,"license":"gemma","categories":null,"tags":["niche"],"acc_tags":[],"description":"The mlabonne/gemma-3-27b-it-abliterated model is a 27 billion parameter instruction-tuned causal language model based on Google's Gemma 3 architecture. This model has been specifically modified using an \"abliteration\" technique to reduce refusals and censorship, aiming for a higher acceptance rate in responses. It is designed for use cases requiring less restrictive content generation while preserving coherence. The model has a context length of 32768 tokens.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":27000000000,"context_length":32768,"created_at":"2025-04-10T02:01:22.783Z","updated_at":"2026-10-15T00:59:47.260Z","hf_created_at":"2025-03-16T19:27:42.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":4860},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["gemma3"],"training":[],"capabilities":[],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["google/gemma-3-27b-it","finetune:google/gemma-3-27b-it"]},"model_class":{"id":"gemma3-27b","created_at":"2025-04-05T04:14:00.000Z","updated_at":"2026-01-27T08:03:30.654Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma3","family_name":"Gemma 3"},"supports_tool_calling":false},{"id":"huihui-ai/Huihui-Qwen3.5-9B-abliterated","name":"huihui-ai/Huihui-Qwen3.5-9B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69ae4e1db51ff114358450af","favorites":155,"downloads":6749,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3.5-9B-abliterated is a 9 billion parameter language model developed by huihui-ai, based on the Qwen3.5-9B architecture. This model has undergone an \"abliteration\" process to significantly reduce its safety filtering and refusal behaviors. It is intended for research and experimental use where the generation of sensitive or controversial content is a known and accepted risk, offering a less censored output compared to its base model.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":true,"isHotRecent":true,"loadingStage":null,"sortRank":100,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":9000000000,"context_length":32768,"created_at":"2026-05-14T20:12:28.808Z","updated_at":"2026-10-14T22:27:26.860Z","hf_created_at":"2026-03-09T04:35:41.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":5258},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3.5-9b","finetune:qwen/qwen3.5-9b"]},"model_class":{"id":"qwen3.5-9b","created_at":"2026-03-08T12:08:53.094Z","updated_at":"2026-03-08T12:09:22.041Z","concurrency_cost":1,"size":9,"parameter_size":9000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen35","family_name":"Qwen 3.5"},"supports_tool_calling":true},{"id":"mlabonne/Meta-Llama-3.1-8B-Instruct-abliterated","name":"mlabonne/Meta-Llama-3.1-8B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"66a175485f58258df71da54d","favorites":216,"downloads":6736,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3.1","categories":null,"tags":null,"acc_tags":[],"description":"mlabonne/Meta-Llama-3.1-8B-Instruct-abliterated is an 8 billion parameter instruction-tuned language model based on Meta's Llama 3.1 architecture, featuring a 32K context length. This model is an uncensored variant created using the abliteration technique, making it suitable for applications requiring less restrictive content generation. It offers a distinct alternative for developers seeking a more open-ended conversational AI.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":true,"isHotRecent":true,"loadingStage":null,"sortRank":100,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":8000000000,"context_length":32768,"created_at":"2024-07-27T10:14:41.537Z","updated_at":"2026-10-15T00:53:55.596Z","hf_created_at":"2024-07-24T21:42:32.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":6996},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["meta-llama/llama-3.1-8b-instruct","finetune:meta-llama/llama-3.1-8b-instruct"]},"model_class":{"id":"llama31-8b","created_at":"2025-10-15T03:53:00.000Z","updated_at":"2026-01-27T07:58:09.462Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama31","family_name":"Llama 3.1"},"supports_tool_calling":true},{"id":"huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2","name":"huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"66f069cd1928e7fe9cd9d206","favorites":66,"downloads":6093,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2 is a 7.6 billion parameter instruction-tuned causal language model, derived from Qwen/Qwen2.5-7B-Instruct. This model has been modified using an 'abliteration' technique to remove censorship, making it suitable for applications requiring unfiltered responses. It maintains a substantial 131,072 token context length and shows competitive performance across various benchmarks, particularly excelling in IF_Eval.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":7600000000,"context_length":32768,"created_at":"2025-02-15T12:20:31.484Z","updated_at":"2026-10-15T00:58:14.345Z","hf_created_at":"2024-09-22T19:02:37.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7100},"classified_tags":{"languages":["zh","en","fr","es","pt","de","it","ru","ja","ko","vi","th","ar"],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen2.5-7b-instruct","finetune:qwen/qwen2.5-7b-instruct"]},"model_class":{"id":"qwen25-7b","created_at":"2025-06-12T09:08:32.442Z","updated_at":"2026-03-09T13:58:41.753Z","concurrency_cost":1,"size":7.6,"parameter_size":7600000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"huihui-ai/Qwen2.5-32B-Instruct-abliterated","name":"huihui-ai/Qwen2.5-32B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"66f94352640009849554e777","favorites":62,"downloads":4813,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"huihui-ai/Qwen2.5-32B-Instruct-abliterated is a 32.8 billion parameter instruction-tuned causal language model, derived from Qwen/Qwen2.5-32B-Instruct. This model has been 'abliterated' to be uncensored, offering a broader range of responses compared to its base model. It supports a 131,072 token context length and is designed for general text generation tasks, particularly in scenarios requiring less restrictive content filtering across multiple languages.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":32799999999.999996,"context_length":32768,"created_at":"2025-02-15T12:21:03.698Z","updated_at":"2026-10-15T00:58:15.552Z","hf_created_at":"2024-09-29T12:08:50.000Z","trending_rank":{"fetched_at":"2026-10-12T07:17:00.029Z","position":4290},"classified_tags":{"languages":["zh","en","fr","es","pt","de","it","ru","ja","ko","vi","th","ar"],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen2.5-32b-instruct","finetune:qwen/qwen2.5-32b-instruct"]},"model_class":{"id":"qwen25-32b","created_at":"2025-06-12T09:08:32.600Z","updated_at":"2026-01-27T07:58:09.463Z","concurrency_cost":2,"size":32.8,"parameter_size":32799999999.999996,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.6-27B-abliterated","name":"huihui-ai/Huihui-Qwen3.6-27B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69ea226dbf690f20d0d58464","favorites":74,"downloads":4635,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3.6-27B-abliterated is a 27 billion parameter language model developed by huihui-ai, based on the Qwen3.6-27B architecture. This model has undergone an 'abliteration' process to significantly reduce its safety filtering and refusal behaviors. It is primarily intended for research and experimental use in scenarios requiring an uncensored model, offering a proof-of-concept for removing refusal mechanisms.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":27000000000,"context_length":32768,"created_at":"2026-06-15T22:03:08.071Z","updated_at":"2026-10-14T22:38:29.676Z","hf_created_at":"2026-04-23T13:45:17.000Z","trending_rank":{"fetched_at":"2026-10-14T02:17:00.030Z","position":9101},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3.6-27b","finetune:qwen/qwen3.6-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Mistral-Small-3.2-24B-Instruct-2506-abliterated","name":"huihui-ai/Huihui-Mistral-Small-3.2-24B-Instruct-2506-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"68655c48e88ec1eb5d827715","favorites":22,"downloads":4471,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Mistral-Small-3.2-24B-Instruct-2506-abliterated is a 24 billion parameter instruction-tuned language model derived from mistralai/Mistral-Small-3.2-24B-Instruct-2506. This model has undergone an \"abliteration\" process to remove refusal behaviors, making it an uncensored version. It maintains a 32768 token context length and is intended for research into uncensored LLM capabilities, particularly for text-based interactions.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":24000000000,"context_length":32768,"created_at":"2025-07-05T03:00:54.385Z","updated_at":"2026-10-15T01:22:29.669Z","hf_created_at":"2025-07-02T16:20:24.000Z","trending_rank":{"fetched_at":"2026-09-02T15:17:00.023Z","position":3670},"classified_tags":{"languages":["en","fr","de","es","pt","it","ja","ko","ru","zh","ar","fa","id","ms","ne","pl","ro","sr","sv","tr","uk","vi","hi","bn"],"tasks":["image-text-to-text","conversational"],"architectures":["mistral3"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["mistralai/mistral-small-3.2-24b-instruct-2506","finetune:mistralai/mistral-small-3.2-24b-instruct-2506"]},"model_class":{"id":"mistral-24b-2503","created_at":"2025-04-26T14:03:41.457Z","updated_at":"2026-01-27T08:03:30.654Z","concurrency_cost":2,"size":24,"parameter_size":24000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"mistral31","family_name":"Mistral 3.1"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated","name":"huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69e2cbf4a8ad7ec14c19f4a4","favorites":98,"downloads":3329,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated model is an uncensored 35.1 billion parameter variant of the Qwen/Qwen3.6-35B-A3B architecture, developed by huihui-ai. This model has undergone \"abliteration\" to significantly reduce its safety filtering and refusal behaviors, making it suitable for research and experimental use cases where content restrictions are undesirable. It is designed for scenarios requiring less constrained output generation, with a context length of 32768 tokens.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":35100000000,"context_length":32768,"created_at":"2026-06-15T22:01:49.141Z","updated_at":"2026-10-14T22:38:29.126Z","hf_created_at":"2026-04-18T00:10:28.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":5353},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_5_moe"],"base_models":["qwen/qwen3.6-35b-a3b","finetune:qwen/qwen3.6-35b-a3b"]},"model_class":{"id":"qwen3_5moe-35b","created_at":"2026-06-13T11:08:45.209Z","updated_at":"2026-06-13T11:08:45.209Z","concurrency_cost":2,"size":35.1,"parameter_size":35100000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3-5-moe","family_name":"Qwen3 5 Moe"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3-4B-abliterated-v2","name":"huihui-ai/Huihui-Qwen3-4B-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"68538643e3278bb954f2f007","favorites":64,"downloads":2827,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3-4B-abliterated-v2 by huihui-ai is a 4 billion parameter uncensored variant of the Qwen3-4B causal language model, featuring a 40960 token context length. This model was created using an abliteration technique to remove refusal behaviors, offering a proof-of-concept for uncensored LLM responses. It is primarily designed for research and experimental use cases where reduced safety filtering is desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":4000000000,"context_length":32768,"created_at":"2025-07-05T04:30:27.332Z","updated_at":"2026-10-15T01:24:18.191Z","hf_created_at":"2025-06-19T03:38:43.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":3255},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-4b","finetune:qwen/qwen3-4b"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3-1.7B-abliterated-v2","name":"huihui-ai/Huihui-Qwen3-1.7B-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"6853916937918576899bd4bf","favorites":5,"downloads":2590,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3-1.7B-abliterated-v2 is a 1.7 billion parameter Qwen3-based causal language model developed by huihui-ai. This model is an uncensored version of Qwen/Qwen3-1.7B, created using an abliteration method to remove refusals. It is optimized for research and experimental use in scenarios where reduced safety filtering is desired, offering improved performance over its predecessor.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":2000000000,"context_length":32768,"created_at":"2026-01-08T07:50:16.335Z","updated_at":"2026-10-14T21:27:48.760Z","hf_created_at":"2025-06-19T04:26:17.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-1.7b","finetune:qwen/qwen3-1.7b"]},"model_class":{"id":"qwen3-1b7","created_at":"2025-06-12T09:08:32.913Z","updated_at":"2026-03-10T07:32:52.904Z","concurrency_cost":1,"size":2,"parameter_size":2000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3-14B-abliterated-v2","name":"huihui-ai/Huihui-Qwen3-14B-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"685189e17be2e22828ae2140","favorites":37,"downloads":2407,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3-14B-abliterated-v2 is a 14 billion parameter Qwen3-based causal language model developed by huihui-ai, featuring a 32768-token context length. This model is specifically engineered to be an uncensored version of Qwen3-14B through an abliteration process, aiming to remove refusal behaviors. It is primarily designed for research and experimental use where reduced safety filtering is desired, offering a proof-of-concept for refusal removal techniques.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":14000000000,"context_length":32768,"created_at":"2025-07-05T04:47:12.048Z","updated_at":"2026-10-15T01:24:42.602Z","hf_created_at":"2025-06-17T15:29:37.000Z","trending_rank":{"fetched_at":"2026-10-02T18:17:00.022Z","position":4772},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-14b","finetune:qwen/qwen3-14b"]},"model_class":{"id":"qwen3-14b","created_at":"2025-05-12T23:45:30.387Z","updated_at":"2026-03-01T08:03:00.238Z","concurrency_cost":1,"size":14,"parameter_size":14000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"wangzhang/Qwen3.5-9B-abliterated","name":"wangzhang/Qwen3.5-9B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69af820e473e186f264b8951","favorites":13,"downloads":2292,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"wangzhang/Qwen3.5-9B-abliterated is a 9 billion parameter causal language model derived from Qwen/Qwen3.5-9B, developed by Wangzhang Wu. This model has been modified using the Abliterix framework to significantly reduce safety-refusal behaviors, achieving a 1% refusal rate while preserving core capabilities. It is optimized for applications requiring an unrestricted language model with practical VRAM requirements and a 32768 token context length.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":9000000000,"context_length":32768,"created_at":"2026-06-15T22:01:35.698Z","updated_at":"2026-10-14T22:38:29.134Z","hf_created_at":"2026-03-10T02:29:34.000Z","trending_rank":{"fetched_at":"2026-09-07T09:17:00.024Z","position":4746},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3_5"],"training":["uncensored","abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["abliterix"],"base_models":["qwen/qwen3.5-9b","finetune:qwen/qwen3.5-9b"]},"model_class":{"id":"qwen3_5-9b","created_at":"2026-06-12T10:59:16.626Z","updated_at":"2026-06-12T10:59:16.626Z","concurrency_cost":1,"size":9,"parameter_size":9000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3-4B-Instruct-2507-abliterated","name":"huihui-ai/Huihui-Qwen3-4B-Instruct-2507-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6894967025b5b79c17de8b27","favorites":45,"downloads":2261,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3-4B-Instruct-2507-abliterated is a 4 billion parameter instruction-tuned causal language model developed by huihui-ai, based on the Qwen3-4B-Instruct-2507 architecture. This model has undergone an \"abliteration\" process to significantly reduce safety filtering and refusals, making it an uncensored variant. It is primarily designed for research and experimental use cases where reduced safety guardrails are desired, offering a proof-of-concept for refusal removal techniques.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":4000000000,"context_length":32768,"created_at":"2025-08-08T04:00:37.697Z","updated_at":"2026-10-15T01:27:33.540Z","hf_created_at":"2025-08-07T12:05:04.000Z","trending_rank":{"fetched_at":"2026-08-15T07:17:00.025Z","position":3628},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-4b-instruct-2507","finetune:qwen/qwen3-4b-instruct-2507"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Llama-3.2-3B-Instruct-abliterated","name":"huihui-ai/Llama-3.2-3B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"66f792029561f0ecc8f55a79","favorites":103,"downloads":2060,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3.2","categories":null,"tags":["niche"],"acc_tags":[],"description":"Llama-3.2-3B-Instruct-abliterated by huihui-ai is a 3.2 billion parameter instruction-tuned causal language model with a 32768 token context length. This model is an uncensored variant of Llama 3.2 3B Instruct, created using the abliteration technique. It maintains or slightly improves performance across several benchmarks like IF_Eval, MMLU Pro, TruthfulQA, BBH, and GPQA, making it suitable for applications requiring an instruction-following model with reduced content restrictions.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":3200000000,"context_length":32768,"created_at":"2025-12-23T17:48:26.714Z","updated_at":"2026-10-14T21:23:34.628Z","hf_created_at":"2024-09-28T05:20:02.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7115},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["meta-llama/llama-3.2-3b-instruct","finetune:meta-llama/llama-3.2-3b-instruct"]},"model_class":{"id":"llama32-3b","created_at":"2025-06-12T19:26:12.601Z","updated_at":"2026-03-10T07:21:03.452Z","concurrency_cost":1,"size":3.2,"parameter_size":3200000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama32","family_name":"Llama 3.2"},"supports_tool_calling":true},{"id":"jenerallee78/gemma-4-26B-A4B-it-ara-abliterated","name":"jenerallee78/gemma-4-26B-A4B-it-ara-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69d41ce8cc1c83107595aef2","favorites":23,"downloads":2005,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"jenerallee78/gemma-4-26B-A4B-it-ara-abliterated is an uncensored 26 billion parameter Gemma 4 model, fine-tuned using Adaptive Refusal Abliteration (ARA) to remove alignment guardrails. This model achieves a significantly lower refusal rate (7.7%) and higher compliance quality (4.6/5) compared to other abliterated versions, while maintaining a low KL divergence from the base model. It is designed for research purposes where compliance with a wider range of prompts is desired, including those the original model would refuse.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":26000000000,"context_length":32768,"created_at":"2026-06-15T16:44:50.149Z","updated_at":"2026-10-15T01:07:09.081Z","hf_created_at":"2026-04-06T20:51:52.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":8897},"classified_tags":{"languages":["ar","en"],"tasks":["image-text-to-text","conversational"],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["moe","multimodal","vision","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["gemma4"],"arxiv":["2402.10260"],"base_models":["google/gemma-4-26b-a4b-it","quantized:google/gemma-4-26b-a4b-it"]},"model_class":{"id":"gemma4-25b","created_at":"2026-06-11T17:32:19.005Z","updated_at":"2026-06-11T17:32:19.005Z","concurrency_cost":2,"size":26,"parameter_size":26000000000,"context_length":32768,"is_moe":true,"active_parameters":4,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"gemma4","family_name":"Gemma 4"},"supports_tool_calling":true},{"id":"KridgeDookie/Qwen3.8-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS","name":"KridgeDookie/Qwen3.8-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a7f72bf9bba3b403fc126cb","favorites":12,"downloads":1937,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"KridgeDookie/Qwen3.8-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS is a 27 billion parameter multimodal derivative of Qwen/Qwen3.8-27B, developed by KridgeDookie. This model is specifically engineered to sharply reduce refusal behavior, achieving 0 refusals across extensive internal prompt screens. It retains the upstream hybrid text-and-vision backbone, making it suitable for applications requiring robust, uncensored responses in both text and multimodal contexts.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-15T11:05:07.808Z","updated_at":"2026-10-15T01:04:38.383Z","hf_created_at":"2026-08-14T19:55:43.000Z","trending_rank":{"fetched_at":"2026-09-26T13:17:00.029Z","position":7662},"classified_tags":{"languages":["multilingual"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen3_5","qwen"],"training":["abliterated","uncensored"],"capabilities":["multimodal","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3.8","refusal-reduction"],"base_models":["qwen/qwen3.8-27b","quantized:qwen/qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"Blackfrost-AI/Qwen3.8-27B-ABLITERATED-BF16","name":"Blackfrost-AI/Qwen3.8-27B-ABLITERATED-BF16","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a7f60bd2933c4e5b219cd36","favorites":32,"downloads":1639,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Blackfrost-AI/Qwen3.8-27B-ABLITERATED-BF16 is a 27 billion parameter dense hybrid Vision-Language Model (VLM) developed by Blackfrost, based on the Qwen3.8 architecture. This model is provided in native BF16 precision and retains vision, reasoning, tool-use, and long-context capabilities. Its primary differentiator is a weight-level modification to significantly reduce false-positive refusals, making it suitable for lawful software engineering and security tasks.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-18T09:32:28.491Z","updated_at":"2026-10-14T23:13:33.141Z","hf_created_at":"2026-08-14T18:38:53.000Z","trending_rank":{"fetched_at":"2026-09-20T17:17:00.021Z","position":7599},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5","qwen"],"training":[],"capabilities":["vision-language","research","red-teaming","code","function-calling","long context","tool-use"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3.8","27b","bf16","dense","derisked","security-research","public-research-preview"],"base_models":["qwen/qwen3.8-27b","finetune:qwen/qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"Goekdeniz-Guelmez/Josiefied-Qwen3-8B-abliterated-v1","name":"Goekdeniz-Guelmez/Josiefied-Qwen3-8B-abliterated-v1","owned_by":"Feather","root_model_id":null,"huggingface_id":"681125494e21a400f6bbedd2","favorites":234,"downloads":1567,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"The Josiefied-Qwen3-8B-abliterated-v1 is an 8 billion parameter language model developed by Gökdeniz Gülmez, based on the Qwen3 architecture with a 32768 token context length. This model is specifically fine-tuned to maximize uncensored behavior and instruction-following abilities, while maintaining tool usage. It is designed for advanced users requiring unrestricted, high-performance language generation, often outperforming its base counterparts on standard benchmarks.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":8000000000,"context_length":32768,"created_at":"2025-05-13T00:00:07.515Z","updated_at":"2026-10-15T01:01:35.746Z","hf_created_at":"2025-04-29T19:15:21.000Z","trending_rank":{"fetched_at":"2026-10-11T19:17:00.030Z","position":4756},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":[],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-8b","finetune:qwen/qwen3-8b"]},"model_class":{"id":"qwen3-8b","created_at":"2025-05-12T23:45:01.939Z","updated_at":"2026-01-27T08:03:30.655Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.5-2B-abliterated","name":"huihui-ai/Huihui-Qwen3.5-2B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69a68a1cae4e7dd37aec734d","favorites":52,"downloads":1504,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-Qwen3.5-2B-abliterated is a 2.3 billion parameter uncensored version of the Qwen/Qwen3.5-2B model, developed by huihui-ai. This model was created using an abliteration technique to remove refusal behaviors, making it suitable for research and experimental use cases where reduced safety filtering is desired. It offers a proof-of-concept for modifying LLM responses without TransformerLens, focusing on generating content without typical safety constraints.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":2300000000,"context_length":32768,"created_at":"2026-07-06T21:48:23.405Z","updated_at":"2026-10-14T22:52:50.878Z","hf_created_at":"2026-03-03T07:13:32.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":5244},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3.5-2b","finetune:qwen/qwen3.5-2b"]},"model_class":{"id":"qwen3_5-2b","created_at":"2026-06-23T14:16:26.914Z","updated_at":"2026-06-23T14:16:26.914Z","concurrency_cost":1,"size":2.3,"parameter_size":2300000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-CyberStrike-OffSec-35B-abliterated","name":"huihui-ai/Huihui-CyberStrike-OffSec-35B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a79fede157e399eeec2c6c4","favorites":104,"downloads":1481,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-CyberStrike-OffSec-35B-abliterated is a 35.1 billion parameter uncensored causal language model, derived from oyildirim/CyberStrike-OffSec-35B. This model has had its refusal mechanisms removed using an abliteration technique, making it suitable for research and experimental use cases where content filtering is intentionally reduced. It offers a 32768 token context length and is designed for scenarios requiring less restrictive content generation.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":35100000000,"context_length":32768,"created_at":"2026-08-11T17:03:42.829Z","updated_at":"2026-10-14T23:09:07.263Z","hf_created_at":"2026-08-10T16:39:58.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":5603},"classified_tags":{"languages":[],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["function-calling","tool-use"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_5_moe","huihui","offensive-security","pentesting","cyberstrike"],"base_models":["oyildirim/cyberstrike-offsec-35b","finetune:oyildirim/cyberstrike-offsec-35b"]},"model_class":{"id":"qwen3_5moe-35b","created_at":"2026-06-13T11:08:45.209Z","updated_at":"2026-06-13T11:08:45.209Z","concurrency_cost":2,"size":35.1,"parameter_size":35100000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3-5-moe","family_name":"Qwen3 5 Moe"},"supports_tool_calling":true},{"id":"mlabonne/Qwen3-14B-abliterated","name":"mlabonne/Qwen3-14B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"68114bf2c9b17d10fd64827c","favorites":55,"downloads":1466,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"mlabonne/Qwen3-14B-abliterated is a 14 billion parameter uncensored version of the Qwen3-14B model, developed by mlabonne. This model utilizes an 'abliteration' technique to remove refusal behaviors, making it suitable for research into latent fine-tuning and refusal mechanisms in large language models. It is designed to achieve a high acceptance rate for various prompts while maintaining coherent output, offering an experimental approach to model modification.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":14000000000,"context_length":32768,"created_at":"2025-07-05T03:45:02.083Z","updated_at":"2026-10-15T01:23:11.037Z","hf_created_at":"2025-04-29T22:00:18.000Z","trending_rank":{"fetched_at":"2026-10-14T06:17:00.030Z","position":7614},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-14b","finetune:qwen/qwen3-14b"]},"model_class":{"id":"qwen3-14b","created_at":"2025-05-12T23:45:30.387Z","updated_at":"2026-03-01T08:03:00.238Z","concurrency_cost":1,"size":14,"parameter_size":14000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"jenerallee78/Qwen3.8-27B-Abliterated-SFT","name":"jenerallee78/Qwen3.8-27B-Abliterated-SFT","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a870f320ddbfc004bca949c","favorites":5,"downloads":1461,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"jenerallee78/Qwen3.8-27B-Abliterated-SFT is a 27 billion parameter Qwen3.8-based model, fine-tuned for refusal removal using an SFT LoRA method. It features a dense hybrid architecture with 64 layers and a 32K context length, retaining vision capabilities. This model is specifically optimized to increase valid fulfillment rates on harmful prompts while preserving termination integrity, making it suitable for applications requiring robust content generation without excessive refusal.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-21T11:13:18.961Z","updated_at":"2026-10-14T23:17:24.837Z","hf_created_at":"2026-08-20T14:29:06.000Z","trending_rank":{"fetched_at":"2026-10-02T20:17:00.021Z","position":7697},"classified_tags":{"languages":[],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen3_5"],"training":["abliterated","sft"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3.8"],"base_models":["qwen/qwen3.8-27b","quantized:qwen/qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-gemma-4-31B-it-abliterated-v2","name":"huihui-ai/Huihui-gemma-4-31B-it-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"69da547f0a7d72741ca1963a","favorites":39,"downloads":1231,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-gemma-4-31B-it-abliterated-v2 is a 31 billion parameter instruction-tuned causal language model, derived from Google's Gemma-4-31B-it. This model has been modified using 'abliteration' techniques to significantly reduce safety filtering and refusals, resulting in a lower perplexity (PPL) value compared to its base model. It is optimized for generating less censored and more direct outputs, making it suitable for research and experimental use cases where content filtering is undesirable.","is_gated":false,"is_private":false,"is_conversational":false,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":31000000000,"context_length":32768,"created_at":"2026-06-15T14:44:20.821Z","updated_at":"2026-10-14T23:52:03.010Z","hf_created_at":"2026-04-11T14:02:39.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":3439},"classified_tags":{"languages":[],"tasks":["image-text-to-text"],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["gemma4","any-to-any"],"base_models":["google/gemma-4-31b-it","finetune:google/gemma-4-31b-it"]},"model_class":{"id":"gemma4-31b","created_at":"2026-04-08T07:52:17.866Z","updated_at":"2026-06-12T21:16:31.087Z","concurrency_cost":2,"size":31,"parameter_size":31000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma4","family_name":"Gemma 4"},"supports_tool_calling":true},{"id":"mlabonne/Qwen3-8B-abliterated","name":"mlabonne/Qwen3-8B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"681144bf3d0b49b7c63f2321","favorites":33,"downloads":1171,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"mlabonne/Qwen3-8B-abliterated is an 8 billion parameter uncensored version of the Qwen/Qwen3-8B causal language model, developed by mlabonne. This model utilizes a novel \"abliteration\" technique to remove refusal behaviors, making it suitable for research into latent fine-tuning and refusal mechanisms in LLMs. It aims to achieve a high acceptance rate for diverse prompts while maintaining coherent output, serving as an experimental tool for understanding model safety and control.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":8000000000,"context_length":32768,"created_at":"2025-07-05T03:41:07.263Z","updated_at":"2026-10-15T01:23:04.925Z","hf_created_at":"2025-04-29T21:29:35.000Z","trending_rank":{"fetched_at":"2026-10-09T11:17:00.029Z","position":4758},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-8b","finetune:qwen/qwen3-8b"]},"model_class":{"id":"qwen3-8b","created_at":"2025-05-12T23:45:01.939Z","updated_at":"2026-01-27T08:03:30.655Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"cyberneurova/CyberNeurova-Gemma-4-12B-it-abliterated","name":"cyberneurova/CyberNeurova-Gemma-4-12B-it-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a25e4a7e37610b0e9ad1459","favorites":1,"downloads":1169,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"CyberNeurova's Gemma-4-12B-it-abliterated is a 12-billion parameter encoder-free trimodal model, based on Google's Gemma 4, with a 32768 token context length. This variant has undergone 'refusal-direction abliteration,' a process that removes safety alignment features from the model's residual stream, making it suitable for cybersecurity research and red-team evaluations. It excels at providing technical information in areas like hacking and cyber-weapons, which are typically refused by safety-aligned models, while preserving core capabilities like coding and reasoning.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027282673},"popularity_level":"3","parameter_size":12000000000,"context_length":32768,"created_at":"2026-07-12T10:17:08.223Z","updated_at":"2026-10-14T22:57:34.099Z","hf_created_at":"2026-06-07T21:37:43.000Z","trending_rank":null,"classified_tags":{"languages":["multilingual","en","es","fr","de","zh","hi","ja","ko","ar","pt","it","ru"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["gemma"],"training":["abliterated","uncensored"],"capabilities":["multimodal","vision-language","research","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["gemma4_unified","gemma-4","audio"],"base_models":["google/gemma-4-12b-it","quantized:google/gemma-4-12b-it"]},"model_class":{"id":"gemma4unified-12b","created_at":"2026-07-10T12:42:57.508Z","updated_at":"2026-07-10T12:42:57.508Z","concurrency_cost":1,"size":12,"parameter_size":12000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma4-unified","family_name":"Gemma4 Unified"},"supports_tool_calling":true},{"id":"huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated","name":"huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6790e7f13b0a6384a42fdaf8","favorites":257,"downloads":1132,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated is a 32.8 billion parameter language model, derived from deepseek-ai/DeepSeek-R1-Distill-Qwen-32B. This model has been modified using an 'abliteration' technique to specifically remove refusal behaviors, making it an uncensored version. It serves as a proof-of-concept for removing LLM refusals without TransformerLens, primarily aimed at use cases requiring direct, unfiltered responses.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":32799999999.999996,"context_length":32768,"created_at":"2025-02-05T12:53:36.326Z","updated_at":"2026-10-15T00:57:57.959Z","hf_created_at":"2025-01-22T12:43:29.000Z","trending_rank":{"fetched_at":"2026-10-09T17:17:00.030Z","position":4508},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["deepseek-ai/deepseek-r1-distill-qwen-32b","finetune:deepseek-ai/deepseek-r1-distill-qwen-32b"]},"model_class":{"id":"qwen25-32b","created_at":"2025-06-12T09:08:32.600Z","updated_at":"2026-01-27T07:58:09.463Z","concurrency_cost":2,"size":32.8,"parameter_size":32799999999.999996,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated","name":"DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"691e6ea1028866c1078d52c3","favorites":61,"downloads":1126,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated is a 1.1 billion parameter Gemma-based instruction-tuned model, uncensored by the Heretic v1.0.1 method. It achieves a refusal rate of 3/100 with a KL divergence of 0.33, significantly reducing censorship compared to the original Gemma 1B-IT's 99/100 refusal rate. This model is optimized for generating content without refusals, while maintaining model integrity for various creative and explicit text generation tasks.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":1000000000,"context_length":32768,"created_at":"2025-11-21T15:12:43.591Z","updated_at":"2026-10-15T01:28:01.137Z","hf_created_at":"2025-11-20T01:28:01.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":2355},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["gemma3_text"],"training":["uncensored","abliterated","finetune"],"capabilities":[],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic"],"base_models":["google/gemma-3-1b-it","finetune:google/gemma-3-1b-it"]},"model_class":{"id":"gemma3t-1b","created_at":"2025-06-28T03:29:29.276Z","updated_at":"2026-01-27T07:55:38.930Z","concurrency_cost":1,"size":1,"parameter_size":1000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma3","family_name":"Gemma 3"},"supports_tool_calling":false},{"id":"huihui-ai/Qwen2.5-0.5B-Instruct-abliterated-v3","name":"huihui-ai/Qwen2.5-0.5B-Instruct-abliterated-v3","owned_by":"Feather","root_model_id":null,"huggingface_id":"680d9c356dc65f5e0bb53ef6","favorites":4,"downloads":1065,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Qwen2.5-0.5B-Instruct-abliterated-v3 is a 0.5 billion parameter instruction-tuned causal language model, based on the Qwen2.5 architecture. Developed by huihui-ai, this model is specifically engineered to be an uncensored version of its base model, achieved through an 'abliteration' process. It excels at responding to a wide range of instructions without refusal, demonstrating a 100% pass rate on a 320-instruction harmful content test set. This model is primarily designed for applications requiring a small, efficient, and unrestricted language model.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":500000000,"context_length":32768,"created_at":"2025-06-16T04:24:59.223Z","updated_at":"2026-10-15T01:21:50.736Z","hf_created_at":"2025-04-27T02:53:41.000Z","trending_rank":{"fetched_at":"2026-08-05T05:17:00.090Z","position":3750},"classified_tags":{"languages":["zh","en","fr","es","pt","de","it","ru","ja","ko","vi","th","ar"],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen2.5-0.5b-instruct","finetune:qwen/qwen2.5-0.5b-instruct"]},"model_class":{"id":"qwen25-0b5","created_at":"2025-06-12T09:08:32.202Z","updated_at":"2026-03-10T07:30:16.141Z","concurrency_cost":1,"size":0.5,"parameter_size":500000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"natong19/Mistral-Nemo-Instruct-2407-abliterated","name":"natong19/Mistral-Nemo-Instruct-2407-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"66bd6b92780d735f179ebd76","favorites":21,"downloads":1029,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Mistral-Nemo-Instruct-2407-abliterated is a 12 billion parameter large language model, jointly trained by Mistral AI and NVIDIA, featuring a 32K context window. This version has undergone ablation to reduce strong refusal directions, while maintaining performance. It is optimized for multilingual and code data, serving as a drop-in replacement for Mistral 7B.","is_gated":false,"is_private":false,"is_conversational":false,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":12000000000,"context_length":32768,"created_at":"2026-01-08T02:27:42.512Z","updated_at":"2026-10-14T21:26:56.750Z","hf_created_at":"2024-08-15T02:44:34.000Z","trending_rank":{"fetched_at":"2026-09-25T04:17:00.786Z","position":4011},"classified_tags":{"languages":["en","fr","de","es","it","pt","ru","zh","ja"],"tasks":[],"architectures":["mistral"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"mistral-nemo","created_at":"1970-01-01T00:00:00.000Z","updated_at":"2026-01-27T07:58:09.462Z","concurrency_cost":1,"size":12,"parameter_size":12000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"mistral","family_name":"Mistral"},"supports_tool_calling":true},{"id":"mlabonne/Qwen3-1.7B-abliterated","name":"mlabonne/Qwen3-1.7B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6810a66a6da28aa30d618596","favorites":16,"downloads":1028,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The mlabonne/Qwen3-1.7B-abliterated model is an uncensored 1.7 billion parameter variant of the Qwen/Qwen3-1.7B architecture, developed by mlabonne. It features a 40960 token context length and is created using an experimental \"abliteration\" technique to remove refusal behaviors. This model is primarily a research project aimed at understanding refusal mechanisms and latent fine-tuning in large language models, offering an uncensored output for specific research or creative applications.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":2000000000,"context_length":32768,"created_at":"2026-01-24T10:31:05.717Z","updated_at":"2026-10-14T21:33:35.156Z","hf_created_at":"2025-04-29T10:14:02.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7583},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-1.7b","finetune:qwen/qwen3-1.7b"]},"model_class":{"id":"qwen3-1b7","created_at":"2025-06-12T09:08:32.913Z","updated_at":"2026-03-10T07:32:52.904Z","concurrency_cost":1,"size":2,"parameter_size":2000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2","name":"huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"67924346bf63cab19d95b4aa","favorites":166,"downloads":1002,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2 is an uncensored version of the DeepSeek-R1-Distill-Qwen-14B model, created by huihui-ai. This 14 billion parameter model focuses on removing refusal behaviors from the original LLM using an abliteration technique. It is designed for use cases requiring direct responses without refusal, serving as a proof-of-concept for refusal removal without TransformerLens.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":14000000000,"context_length":32768,"created_at":"2025-02-15T12:09:54.335Z","updated_at":"2026-10-15T00:58:06.865Z","hf_created_at":"2025-01-23T13:25:26.000Z","trending_rank":{"fetched_at":"2026-10-10T00:17:00.027Z","position":4509},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["deepseek-ai/deepseek-r1-distill-qwen-14b","finetune:deepseek-ai/deepseek-r1-distill-qwen-14b"]},"model_class":{"id":"qwen2-14b-lc","created_at":"2025-02-15T03:15:00.000Z","updated_at":"2025-02-15T03:15:00.000Z","concurrency_cost":1,"size":14,"parameter_size":14000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen2","family_name":"Qwen 2"},"supports_tool_calling":true},{"id":"0arch-io/dolphin-v2-8b-abliterated","name":"0arch-io/dolphin-v2-8b-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"699e0d21c77361846059dec5","favorites":4,"downloads":973,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Dolphin V2 8B Abliterated is an 8.2 billion parameter uncensored language model built on Qwen3-8B, fine-tuned on 1.35 million high-quality instruction samples. Developed by 0arch-io for TRC research, it features a 40960 token maximum context length and has undergone an 'abliteration' process to remove refusal behaviors. This model is designed to comply with any request without refusing, making it suitable for research into uncensored model capabilities.","is_gated":false,"is_private":false,"is_conversational":false,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":8000000000,"context_length":32768,"created_at":"2026-02-27T11:45:42.969Z","updated_at":"2026-10-14T21:48:56.750Z","hf_created_at":"2026-02-24T20:42:09.000Z","trending_rank":{"fetched_at":"2026-10-07T14:17:00.023Z","position":5546},"classified_tags":{"languages":["en"],"tasks":["text-generation"],"architectures":["qwen3","dolphin"],"training":["uncensored","abliterated","sft"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["trc"],"arxiv":["2406.11717"],"base_models":["qwen/qwen3-8b","quantized:qwen/qwen3-8b"]},"model_class":{"id":"qwen3-8b","created_at":"2025-05-12T23:45:01.939Z","updated_at":"2026-01-27T08:03:30.655Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"TobiasLogic/Qwen2.5-Coder-32B-abliterated","name":"TobiasLogic/Qwen2.5-Coder-32B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a4531ffba8446a30fd1cb7b","favorites":10,"downloads":903,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"TobiasLogic/Qwen2.5-Coder-32B-abliterated is an uncensored, 32-billion parameter Qwen2.5-Coder-32B-Instruct derivative model. It has been modified to remove refusal behaviors by orthogonalizing the refusal direction out of residual-writing weights, resulting in a 0.0% refusal rate on harmful prompts. This static weight edit maintains the base model's strong coding abilities, making it suitable for local coding assistance and agent work requiring uncensored responses.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":32799999999.999996,"context_length":32768,"created_at":"2026-07-03T13:17:15.972Z","updated_at":"2026-10-14T22:48:49.310Z","hf_created_at":"2026-07-01T15:27:59.000Z","trending_rank":{"fetched_at":"2026-10-02T08:17:00.021Z","position":6909},"classified_tags":{"languages":["en"],"tasks":["text-generation","conversational"],"architectures":["qwen2","qwen2.5"],"training":["abliterated","uncensored"],"capabilities":["code","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"arxiv":["2409.12186"],"base_models":["qwen/qwen2.5-coder-32b-instruct","finetune:qwen/qwen2.5-coder-32b-instruct"],"doi":["10.57967/hf/9534"]},"model_class":{"id":"qwen25-32b","created_at":"2025-06-12T09:08:32.600Z","updated_at":"2026-01-27T07:58:09.463Z","concurrency_cost":2,"size":32.8,"parameter_size":32799999999.999996,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated","name":"huihui-ai/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69e73e51d20baec02ef7f768","favorites":221,"downloads":900,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated model is a 35.1 billion parameter language model based on the Qwen3.6 architecture. It is an uncensored version of lordx64/Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled, created using an abliteration technique to remove safety filtering. This model is designed for research and experimental use where the removal of refusal behaviors is desired, offering outputs without typical safety constraints.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":true,"isHotRecent":true,"loadingStage":null,"sortRank":100,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":35100000000,"context_length":32768,"created_at":"2026-06-15T21:06:25.819Z","updated_at":"2026-10-14T22:38:20.485Z","hf_created_at":"2026-04-21T09:07:29.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":9046},"classified_tags":{"languages":["en"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen"],"training":["distillation","lora","unsloth","abliterated","uncensored"],"capabilities":["reasoning","chain-of-thought","moe","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_5_moe","qwen3.6","mixture-of-experts"],"base_models":["lordx64/qwen3.6-35b-a3b-claude-4.7-opus-reasoning-distilled","adapter:lordx64/qwen3.6-35b-a3b-claude-4.7-opus-reasoning-distilled"]},"model_class":{"id":"qwen3_5moe-35b","created_at":"2026-06-13T11:08:45.209Z","updated_at":"2026-06-13T11:08:45.209Z","concurrency_cost":2,"size":35.1,"parameter_size":35100000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3-5-moe","family_name":"Qwen3 5 Moe"},"supports_tool_calling":true},{"id":"mlabonne/Qwen3-4B-abliterated","name":"mlabonne/Qwen3-4B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"681142e366e75cf561937e5f","favorites":20,"downloads":876,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"mlabonne/Qwen3-4B-abliterated is a 4 billion parameter Qwen3-based language model developed by mlabonne, featuring a 40960 token context length. This model is an uncensored variant created using a novel \"abliteration\" technique to remove refusal behaviors. It serves as a research project to explore latent fine-tuning and the removal of censorship in LLMs, aiming for high acceptance rates while maintaining coherent output.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":4000000000,"context_length":32768,"created_at":"2025-07-05T03:49:27.156Z","updated_at":"2026-10-15T01:23:17.236Z","hf_created_at":"2025-04-29T21:21:39.000Z","trending_rank":{"fetched_at":"2026-06-29T14:17:00.024Z","position":3428},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-4b","finetune:qwen/qwen3-4b"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated","name":"huihui-ai/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a3dde9e90b447c0205690ce","favorites":31,"downloads":875,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwythos-9B-Claude-Mythos-5-1M-abliterated is a 9 billion parameter language model, an uncensored variant of empero-ai/Qwythos-9B-Claude-Mythos-5-1M, featuring a 32768 token context length. This model is specifically designed to strictly adhere to initial prompt instructions in non-thinking mode, delivering only the requested output without conversational filler. Its primary differentiator is the significant reduction in safety filtering, making it suitable for research and experimental use where unconstrained content generation is desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":9000000000,"context_length":32768,"created_at":"2026-06-28T17:11:57.166Z","updated_at":"2026-10-14T22:47:12.006Z","hf_created_at":"2026-06-26T02:06:22.000Z","trending_rank":{"fetched_at":"2026-10-02T21:17:00.021Z","position":6815},"classified_tags":{"languages":["en"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen3_5"],"training":["uncensored","sft","abliterated"],"capabilities":["reasoning","long context","function-calling","tool-use","agent"],"modalities":["vision"],"domains":["cybersecurity","biomedical"],"creative":[],"content_flags":[],"other":["qwen3.5","1m-context","full-fine-tune"],"base_models":["empero-ai/qwythos-9b-claude-mythos-5-1m","finetune:empero-ai/qwythos-9b-claude-mythos-5-1m"]},"model_class":{"id":"qwen3_5-9b","created_at":"2026-06-12T10:59:16.626Z","updated_at":"2026-06-12T10:59:16.626Z","concurrency_cost":1,"size":9,"parameter_size":9000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"DavidAU/Qwen3-0.6B-heretic-abliterated-uncensored","name":"DavidAU/Qwen3-0.6B-heretic-abliterated-uncensored","owned_by":"Feather","root_model_id":null,"huggingface_id":"691e8e4b028866c107902afe","favorites":47,"downloads":866,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"DavidAU/Qwen3-0.6B-heretic-abliterated-uncensored is a 0.8 billion parameter Qwen3-based language model, uncensored using the Heretic v1.0.1 method. It achieves a refusal rate of 6/100 and a KL divergence of 0.00, indicating minimal damage to the model's original state while significantly reducing censorship. This model is primarily designed for users seeking a less restrictive AI for content generation, particularly for creative writing and roleplay scenarios that might involve sensitive or explicit themes.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":800000000,"context_length":32768,"created_at":"2026-02-25T08:42:40.610Z","updated_at":"2026-10-14T21:48:29.643Z","hf_created_at":"2025-11-20T03:43:07.000Z","trending_rank":{"fetched_at":"2026-09-24T18:17:00.081Z","position":5329},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["uncensored","abliterated","finetune"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic"],"base_models":["qwen/qwen3-0.6b","finetune:qwen/qwen3-0.6b"]},"model_class":{"id":"qwen3-0b6","created_at":"2025-06-12T09:08:32.830Z","updated_at":"2026-03-10T07:32:21.140Z","concurrency_cost":1,"size":0.8,"parameter_size":800000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"Goekdeniz-Guelmez/Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1","name":"Goekdeniz-Guelmez/Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1","owned_by":"Feather","root_model_id":null,"huggingface_id":"66ede09c3b250e9eca7aba91","favorites":6,"downloads":852,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Goekdeniz-Guelmez/Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1 is a 1.5 billion parameter Qwen2.5-based instruction-tuned language model developed by Gökdeniz Gülmez. This model is a further fine-tuned version of Qwen/Qwen2.5-1.5B-Instruct, specifically modified for uncensored responses. It is designed to act as an \"Outstandingly Smart Intelligent Entity\" (J.O.S.I.E.) with all refusal vectors removed, making it suitable for use cases requiring unrestricted AI assistance. The model supports a context length of 32768 tokens and is multilingual, covering languages such as English, German, and others.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":1500000000,"context_length":32768,"created_at":"2026-05-07T09:52:46.579Z","updated_at":"2026-10-14T22:19:34.922Z","hf_created_at":"2024-09-20T20:52:44.000Z","trending_rank":{"fetched_at":"2026-05-06T22:00:00.135Z","position":915485},"classified_tags":{"languages":["zh","en","fr","es","pt","de","it","ru","ja","ko","vi","th","ar"],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":[],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen2.5-1.5b","finetune:qwen/qwen2.5-1.5b"]},"model_class":{"id":"qwen2-1b5","created_at":"2025-06-12T09:08:31.819Z","updated_at":"2026-03-10T07:29:28.502Z","concurrency_cost":1,"size":1.5,"parameter_size":1500000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen2","family_name":"Qwen 2"},"supports_tool_calling":true},{"id":"failspy/Llama-3-8B-Instruct-abliterated","name":"failspy/Llama-3-8B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"663a52ade14047f710a1ea9f","favorites":13,"downloads":843,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"llama3","categories":null,"tags":["niche"],"acc_tags":[],"description":"failspy/Llama-3-8B-Instruct-abliterated is an 8 billion parameter instruction-tuned Llama 3 model, derived from meta-llama/Llama-3-8B-Instruct. This variant has undergone specific weight manipulation to inhibit refusal behaviors, based on research into refusal mediation. It maintains the original Llama 3 instruction tuning while exploring the effects of orthogonalizing refusal directions, making it suitable for applications where reduced refusal tendencies are desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":8000000000,"context_length":8192,"created_at":"2024-06-20T19:14:05.514Z","updated_at":"2026-10-15T00:48:03.139Z","hf_created_at":"2024-05-07T16:11:25.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":6891},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"llama3-8b","created_at":"2025-12-02T14:43:00.000Z","updated_at":"2026-03-06T06:13:02.721Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama3","family_name":"Llama 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-ThinkingCap-Qwen3.6-27B-abliterated","name":"huihui-ai/Huihui-ThinkingCap-Qwen3.6-27B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a539c6e6ffc2ee306c798db","favorites":63,"downloads":833,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-ThinkingCap-Qwen3.6-27B-abliterated is a 27 billion parameter language model derived from bottlecapai/ThinkingCap-Qwen3.6-27B, specifically modified by huihui-ai to significantly reduce safety filtering. This model has undergone an \"abliteration\" process to remove refusal behaviors, making it suitable for research and experimental use cases where content filtering is intentionally minimized. It features a 32768 token context length and is designed for scenarios requiring less constrained output generation.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":27000000000,"context_length":32768,"created_at":"2026-07-13T12:02:52.080Z","updated_at":"2026-10-14T22:57:54.135Z","hf_created_at":"2026-07-12T13:53:50.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":9837},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_6","token-efficient","efficient-thinking","huihui"],"base_models":["bottlecapai/thinkingcap-qwen3.6-27b","finetune:bottlecapai/thinkingcap-qwen3.6-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/DeepSeek-R1-Distill-Llama-70B-abliterated","name":"huihui-ai/DeepSeek-R1-Distill-Llama-70B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"679746018c7a5e66d4d2d0a5","favorites":104,"downloads":820,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/DeepSeek-R1-Distill-Llama-70B-abliterated model is an uncensored 70 billion parameter language model based on the DeepSeek-R1-Distill-Llama architecture. It has been modified using an 'abliteration' technique to remove refusal behaviors, making it suitable for applications requiring direct responses without censorship. This model is designed for users who need an LLM that provides straightforward answers, even to potentially sensitive queries, by mitigating built-in refusal mechanisms.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":70000000000,"context_length":32768,"created_at":"2025-01-30T00:00:40.850Z","updated_at":"2026-10-15T00:57:53.698Z","hf_created_at":"2025-01-27T08:38:25.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7345},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["deepseek-ai/deepseek-r1-distill-llama-70b","finetune:deepseek-ai/deepseek-r1-distill-llama-70b"]},"model_class":{"id":"llama31-70b","created_at":"2025-10-15T03:53:00.000Z","updated_at":"2026-01-27T07:58:09.462Z","concurrency_cost":4,"size":70,"parameter_size":70000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama31","family_name":"Llama 3.1"},"supports_tool_calling":true},{"id":"DavidAU/Qwen3-4B-2507-Thinking-heretic-abliterated-uncensored","name":"DavidAU/Qwen3-4B-2507-Thinking-heretic-abliterated-uncensored","owned_by":"Feather","root_model_id":null,"huggingface_id":"693a68f111d91756faf7b8eb","favorites":12,"downloads":790,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"DavidAU/Qwen3-4B-2507-Thinking-heretic-abliterated-uncensored is a 4 billion parameter Qwen3-based language model, developed by DavidAU, featuring a 256k context length. This model has been specifically de-censored using the Heretic v1.0.1 method, achieving a significantly reduced refusal rate of 4/100 while maintaining a low KL divergence of 0.05. It is optimized for generating content without refusals, making it suitable for applications requiring unrestricted text generation.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":4000000000,"context_length":32768,"created_at":"2025-12-17T22:22:04.494Z","updated_at":"2026-10-14T21:22:45.059Z","hf_created_at":"2025-12-11T06:47:13.000Z","trending_rank":{"fetched_at":"2026-08-15T10:17:00.024Z","position":4032},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["uncensored","abliterated","finetune"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic"],"base_models":["qwen/qwen3-4b-thinking-2507","finetune:qwen/qwen3-4b-thinking-2507"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"andrewting/Swift-Qwen3.8-27B-Abliterated","name":"andrewting/Swift-Qwen3.8-27B-Abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6aaeae9105f6ef9717c57c07","favorites":0,"downloads":703,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"other","categories":null,"tags":["niche"],"acc_tags":[],"description":"andrewting/Swift-Qwen3.8-27B-Abliterated is an experimental 27 billion parameter BF16 derivative of ukisai/Swift-Qwen3.8-27b, developed by Andrew Ting. This model features a reconstructed rank-6 refusal subspace, significantly reducing refusal-phrase rates while maintaining task completion. It is designed for applications requiring a reduction in soft refusals, though it is not refusal-free and exhibits material distribution drift on ordinary prompts.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":27000000000,"context_length":32768,"created_at":"2026-09-21T11:13:44.298Z","updated_at":"2026-10-14T23:37:38.183Z","hf_created_at":"2026-09-19T15:47:29.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated"],"capabilities":["reasoning","vision-language","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_8","refusal-reduced"],"base_models":["ukisai/swift-qwen3.8-27b","finetune:ukisai/swift-qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"hotdogs/Qwen3.8-27B-abliterated","name":"hotdogs/Qwen3.8-27B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a802b1668ed361e9967c72b","favorites":10,"downloads":695,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The hotdogs/Qwen3.8-27B-abliterated model is a training-free abliterated version of Qwen/Qwen3.8-27B, a dense 27B native vision-language model with a hybrid full-attention and linear-attention architecture. This model has undergone a forward-only weight edit to remove refusal behavior by orthogonalizing a single \"refusal direction\" from the residual stream, without fine-tuning or data poisoning. It is designed for alignment and safety research, allowing for red-teaming and studying refusal mechanics while largely preserving the base model's capabilities.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-16T12:48:43.355Z","updated_at":"2026-10-15T00:41:20.152Z","hf_created_at":"2026-08-15T09:02:14.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":2481},"classified_tags":{"languages":["en","zh"],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3.8-27b","finetune:qwen/qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"mlabonne/Qwen3-0.6B-abliterated","name":"mlabonne/Qwen3-0.6B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6810a1cf62c2996ef8337ede","favorites":15,"downloads":648,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The mlabonne/Qwen3-0.6B-abliterated is an 0.8 billion parameter causal language model based on the Qwen3 architecture, developed by mlabonne. This model is an uncensored version of Qwen/Qwen3-0.6B, created using a novel abliteration technique. It serves as a research project to explore refusal mechanisms and latent fine-tuning in LLMs, aiming to achieve a high acceptance rate for diverse outputs while maintaining coherence. Its primary differentiator is its experimental abliteration to remove censorship, making it suitable for research into model safety and control.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":800000000,"context_length":32768,"created_at":"2026-01-25T17:00:23.455Z","updated_at":"2026-10-14T21:37:48.582Z","hf_created_at":"2025-04-29T09:54:23.000Z","trending_rank":{"fetched_at":"2026-09-07T12:17:00.023Z","position":3639},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-0.6b","finetune:qwen/qwen3-0.6b"]},"model_class":{"id":"qwen3-0b6","created_at":"2025-06-12T09:08:32.830Z","updated_at":"2026-03-10T07:32:21.140Z","concurrency_cost":1,"size":0.8,"parameter_size":800000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"OliviaRossi/MiMo-Ornith-9B-AGSI-Abliterated-HQ","name":"OliviaRossi/MiMo-Ornith-9B-AGSI-Abliterated-HQ","owned_by":"Feather","root_model_id":null,"huggingface_id":"6ab4eff9dbeada22faaaec37","favorites":6,"downloads":639,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"OliviaRossi/MiMo-Ornith-9B-AGSI-Abliterated-HQ is a 9.7 billion parameter Qwen3_5-based model, developed by OliviaRossi, engineered for high-precision reasoning and agentic capabilities. It uniquely combines chain-of-thought reasoning from MiMo-V2.6 with reinforcement-learned terminal execution from Ornith-1.5, using Adaptive Geodesic Spectral Interpolation (AGSI) for merging. The model features Bayesian Representation Abliteration (Abliterix) to remove refusal vectors while preserving reasoning, making it suitable for complex algorithmic synthesis, coding, and multi-step problem-solving without censorship.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":9000000000,"context_length":32768,"created_at":"2026-09-26T10:28:23.881Z","updated_at":"2026-10-14T23:38:59.418Z","hf_created_at":"2026-09-24T09:40:09.000Z","trending_rank":{"fetched_at":"2026-10-12T06:17:00.028Z","position":8784},"classified_tags":{"languages":["en","zh"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen3_5"],"training":["merge","abliterated","uncensored"],"capabilities":["reasoning","code","agent","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["agsi","abliterix","terminal-use"],"base_models":["xiaomimimo/mimo-v2.6-distill-qwen-9b","merge:xiaomimimo/mimo-v2.6-distill-qwen-9b","ornith-ai/ornith-1.5-9b","merge:ornith-ai/ornith-1.5-9b"]},"model_class":{"id":"qwen3_5-9b","created_at":"2026-06-12T10:59:16.626Z","updated_at":"2026-06-12T10:59:16.626Z","concurrency_cost":1,"size":9,"parameter_size":9000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/Qwen2.5-72B-Instruct-abliterated","name":"huihui-ai/Qwen2.5-72B-Instruct-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"671c59590cd14bd7351e05ca","favorites":84,"downloads":631,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"other","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Qwen2.5-72B-Instruct-abliterated model is a 72.7 billion parameter instruction-tuned causal language model, derived from Qwen/Qwen2.5-72B-Instruct. This version has been specifically modified using 'abliteration' techniques to remove refusal behaviors, offering an uncensored response capability. It is designed for applications requiring direct answers without built-in content restrictions, making it suitable for research into refusal removal and specific use cases where unconstrained output is desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":true,"isHotRecent":true,"loadingStage":null,"sortRank":100,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":72700000000,"context_length":32768,"created_at":"2024-12-07T00:10:35.199Z","updated_at":"2026-10-15T00:57:15.780Z","hf_created_at":"2024-10-26T02:52:09.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7160},"classified_tags":{"languages":["zh","en","fr","es","pt","de","it","ru","ja","ko","vi","th","ar"],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen2.5-72b-instruct","finetune:qwen/qwen2.5-72b-instruct"]},"model_class":{"id":"qwen25-72b","created_at":"2025-06-12T09:08:32.684Z","updated_at":"2025-12-05T19:49:46.502Z","concurrency_cost":4,"size":72.7,"parameter_size":72700000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"huihui-ai/DeepSeek-R1-Distill-Qwen-7B-abliterated-v2","name":"huihui-ai/DeepSeek-R1-Distill-Qwen-7B-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"679c416d9490a955a848438b","favorites":56,"downloads":621,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/DeepSeek-R1-Distill-Qwen-7B-abliterated-v2 is a 7.6 billion parameter language model, derived from deepseek-ai/DeepSeek-R1-Distill-Qwen-7B, with a 131,072 token context length. This version has been modified using an abliteration technique to remove refusal behaviors, making it an uncensored variant. It is designed for use cases requiring direct responses without built-in content restrictions, serving as a proof-of-concept for refusal removal without TransformerLens.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":7600000000,"context_length":32768,"created_at":"2025-02-15T12:10:08.845Z","updated_at":"2026-10-15T00:58:07.520Z","hf_created_at":"2025-01-31T03:20:13.000Z","trending_rank":{"fetched_at":"2026-09-09T23:17:00.023Z","position":3458},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["deepseek-ai/deepseek-r1-distill-qwen-7b","finetune:deepseek-ai/deepseek-r1-distill-qwen-7b"]},"model_class":{"id":"qwen2-7b","created_at":"2025-06-12T09:08:31.903Z","updated_at":"2026-03-09T13:58:05.398Z","concurrency_cost":1,"size":7.6,"parameter_size":7600000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen2","family_name":"Qwen 2"},"supports_tool_calling":true},{"id":"3MPER0RR/Ornith-1.5-35B-A3B-3MPER0RR-abliterated","name":"3MPER0RR/Ornith-1.5-35B-A3B-3MPER0RR-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6aaa8a2bbeed36f2ddbc652e","favorites":0,"downloads":576,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"mit","categories":null,"tags":["niche"],"acc_tags":[],"description":"3MPER0RR/Ornith-1.5-35B-A3B-3MPER0RR-abliterated is a 35.1 billion parameter language model developed by 3MPER0RR, based on the Huihui-Ornith-1.5-35B-A3B-abliterated base model. This model has undergone a multi-round abliteration process, indicating a focus on refinement and optimization. It is designed for research and experimentation, offering a foundation for various natural language processing tasks.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":35100000000,"context_length":32768,"created_at":"2026-09-25T18:46:55.737Z","updated_at":"2026-10-14T23:38:56.130Z","hf_created_at":"2026-09-16T12:23:07.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":[],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_5_moe"],"base_models":["huihui-ai/huihui-ornith-1.5-35b-a3b-abliterated","finetune:huihui-ai/huihui-ornith-1.5-35b-a3b-abliterated"]},"model_class":{"id":"qwen3_5moe-35b","created_at":"2026-06-13T11:08:45.209Z","updated_at":"2026-06-13T11:08:45.209Z","concurrency_cost":2,"size":35.1,"parameter_size":35100000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3-5-moe","family_name":"Qwen3 5 Moe"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated","name":"huihui-ai/Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a373a5134487f4a354fe050","favorites":167,"downloads":564,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated is a 12 billion parameter Gemma-4 based model, derived from yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1, with a 32768 token context length. This version has undergone 'abliteration' to significantly reduce safety filtering and refusal behaviors. It is intended for research and experimental use where uncensored outputs are specifically desired, rather than for production or public-facing applications.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"5","parameter_size":12000000000,"context_length":32768,"created_at":"2026-07-11T23:33:45.040Z","updated_at":"2026-10-14T22:57:32.952Z","hf_created_at":"2026-06-21T01:11:45.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":9605},"classified_tags":{"languages":[],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["code","reasoning","thinking","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["gemma4_unified","gemma4"],"base_models":["yuxinlu1/gemma-4-12b-coder-fable5-composer2.5-v1","finetune:yuxinlu1/gemma-4-12b-coder-fable5-composer2.5-v1"]},"model_class":{"id":"gemma4unified-12b","created_at":"2026-07-10T12:42:57.508Z","updated_at":"2026-07-10T12:42:57.508Z","concurrency_cost":1,"size":12,"parameter_size":12000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma4-unified","family_name":"Gemma4 Unified"},"supports_tool_calling":true},{"id":"RootMonsteR/Qwen3-14B-Abliterated","name":"RootMonsteR/Qwen3-14B-Abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a0c3f64b34b5e612ecce1ba","favorites":2,"downloads":552,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"RootMonsteR/Qwen3-14B-Abliterated is a 14.8 billion parameter causal language model developed by RootMonsteR, based on Qwen/Qwen3-14B. This variant is specifically engineered to suppress refusal behaviors while preserving core capabilities, making it highly suitable for autonomous agents, tool-use workflows, and authorized security research. It maintains Qwen3-14B's reasoning, coding, and tool-calling capabilities with a very low KL divergence of 0.0333 from the base model, while reducing measured refusals by approximately 90%.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":14000000000,"context_length":32768,"created_at":"2026-06-01T10:04:31.733Z","updated_at":"2026-10-14T22:33:45.875Z","hf_created_at":"2026-05-19T10:45:56.000Z","trending_rank":{"fetched_at":"2026-08-18T19:17:00.037Z","position":5340},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling","agent","reasoning"],"modalities":["text"],"domains":["security"],"creative":[],"content_flags":[],"other":["heretic","heretic-llm","qwen3-14b","reproducible"],"arxiv":["2505.09388"],"base_models":["qwen/qwen3-14b","finetune:qwen/qwen3-14b"]},"model_class":{"id":"qwen3-14b","created_at":"2025-05-12T23:45:30.387Z","updated_at":"2026-03-01T08:03:00.238Z","concurrency_cost":1,"size":14,"parameter_size":14000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"sci4ai/Qwen2.5-Coder-14B-Abliterated","name":"sci4ai/Qwen2.5-Coder-14B-Abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69c938b0fc5631a422d216b5","favorites":0,"downloads":536,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"Qwen2.5-Coder-14B-Abliterated is a 14.8 billion parameter language model developed by sci4ai, based on the Qwen2.5-Coder-14B-Instruct architecture with a 32768 token context length. This model has undergone 'abliteration' to remove refusal behaviors, specifically targeting code-related safety guardrails. It is optimized for code generation tasks, providing responses that the original model would typically refuse.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":14800000000,"context_length":32768,"created_at":"2026-06-26T15:41:45.100Z","updated_at":"2026-10-14T22:44:27.568Z","hf_created_at":"2026-03-29T14:35:28.000Z","trending_rank":null,"classified_tags":{"languages":["en"],"tasks":["text-generation","conversational"],"architectures":["qwen2","qwen2.5"],"training":["abliterated","uncensored"],"capabilities":["code","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen2.5-coder-14b-instruct","finetune:qwen/qwen2.5-coder-14b-instruct"]},"model_class":{"id":"qwen25-14b","created_at":"2025-06-12T09:08:32.518Z","updated_at":"2026-01-27T07:58:09.463Z","concurrency_cost":1,"size":14.8,"parameter_size":14800000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"huihui-ai/Qwen3-32B-abliterated","name":"huihui-ai/Qwen3-32B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"68428f58ba3e4017ffead659","favorites":40,"downloads":534,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Qwen3-32B-abliterated model is a 32 billion parameter causal language model derived from Qwen/Qwen3-32B, specifically modified to remove refusal behaviors. Developed by huihui-ai, this model utilizes an abliteration technique for uncensoring, offering a proof-of-concept for refusal removal without TransformerLens. It is designed for applications requiring an uncensored large language model, particularly for research into model safety and control mechanisms.","is_gated":true,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":32000000000,"context_length":32768,"created_at":"2025-06-13T16:55:01.660Z","updated_at":"2026-10-15T01:31:46.929Z","hf_created_at":"2025-06-06T06:48:56.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7694},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-32b","finetune:qwen/qwen3-32b"]},"model_class":{"id":"qwen3-32b","created_at":"2025-05-12T23:46:02.728Z","updated_at":"2026-01-27T07:29:53.291Z","concurrency_cost":2,"size":32,"parameter_size":32000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Qwen2.5-3B-Instruct-abliterated-SFT","name":"huihui-ai/Qwen2.5-3B-Instruct-abliterated-SFT","owned_by":"Feather","root_model_id":null,"huggingface_id":"67fb202fc590bfb7309508ae","favorites":3,"downloads":531,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"huihui-ai/Qwen2.5-3B-Instruct-abliterated-SFT is an instruction-tuned language model developed by huihui-ai, fine-tuned from the Qwen2.5-3B-Instruct-abliterated base model. This model is specifically trained on the Guilherme34_uncensor dataset, focusing on uncensored content generation. It is designed for conversational AI applications requiring less restrictive content filtering, leveraging the Qwen2.5 architecture.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":3100000000,"context_length":32768,"created_at":"2025-07-05T03:57:58.303Z","updated_at":"2026-10-15T01:23:34.227Z","hf_created_at":"2025-04-13T02:23:43.000Z","trending_rank":null,"classified_tags":{"languages":["en"],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["unsloth","abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"datasets":["huihui-ai/guilherme34_uncensor"],"base_models":["huihui-ai/qwen2.5-3b-instruct-abliterated","finetune:huihui-ai/qwen2.5-3b-instruct-abliterated"]},"model_class":{"id":"qwen25-3b","created_at":"2025-06-12T09:08:32.360Z","updated_at":"2026-03-10T07:31:51.709Z","concurrency_cost":1,"size":3.1,"parameter_size":3100000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"huihui-ai/DeepSeek-R1-Distill-Llama-8B-abliterated","name":"huihui-ai/DeepSeek-R1-Distill-Llama-8B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6790ed1f28b3a471d9ddeb53","favorites":79,"downloads":493,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/DeepSeek-R1-Distill-Llama-8B-abliterated is an 8 billion parameter language model, derived from deepseek-ai/DeepSeek-R1-Distill-Llama-8B, with a 32768 token context length. This model has been modified using 'abliteration' techniques to remove refusal behaviors, making it an uncensored variant. It is primarily designed for use cases requiring direct responses without content filtering, serving as a proof-of-concept for refusal removal without TransformerLens.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":8000000000,"context_length":32768,"created_at":"2025-01-27T00:00:37.127Z","updated_at":"2026-10-15T00:57:51.850Z","hf_created_at":"2025-01-22T13:05:35.000Z","trending_rank":{"fetched_at":"2026-08-10T08:17:00.022Z","position":3457},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["llama"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["deepseek-ai/deepseek-r1-distill-llama-8b","finetune:deepseek-ai/deepseek-r1-distill-llama-8b"]},"model_class":{"id":"llama31-8b","created_at":"2025-10-15T03:53:00.000Z","updated_at":"2026-01-27T07:58:09.462Z","concurrency_cost":1,"size":8,"parameter_size":8000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"llama31","family_name":"Llama 3.1"},"supports_tool_calling":true},{"id":"huihui-ai/DeepSeek-R1-Distill-Qwen-1.5B-abliterated","name":"huihui-ai/DeepSeek-R1-Distill-Qwen-1.5B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"67f9f6f36e9f2d8ad944b06f","favorites":7,"downloads":491,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/DeepSeek-R1-Distill-Qwen-1.5B-abliterated is a 1.5 billion parameter language model, post-trained by huihui-ai, based on deepseek-ai's DeepSeek-R1-Distill-Qwen-1.5B. This model has been specifically modified to be uncensored through an ablation fine-tuning process. With a substantial 131,072 token context length, it is designed for reasoning tasks where unconstrained responses are desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":1500000000,"context_length":32768,"created_at":"2026-01-25T16:37:17.246Z","updated_at":"2026-10-14T21:37:46.289Z","hf_created_at":"2025-04-12T05:15:31.000Z","trending_rank":{"fetched_at":"2026-09-22T06:17:00.021Z","position":4385},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["unsloth","abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["deepseek-ai/deepseek-r1-distill-qwen-1.5b","finetune:deepseek-ai/deepseek-r1-distill-qwen-1.5b"]},"model_class":{"id":"qwen2-1b5","created_at":"2025-06-12T09:08:31.819Z","updated_at":"2026-03-10T07:29:28.502Z","concurrency_cost":1,"size":1.5,"parameter_size":1500000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen2","family_name":"Qwen 2"},"supports_tool_calling":true},{"id":"IlyaGusev/gemma-2-2b-it-abliterated","name":"IlyaGusev/gemma-2-2b-it-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"66aa7e70b712a683f6c218e3","favorites":57,"downloads":450,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"gemma","categories":null,"tags":["niche"],"acc_tags":[],"description":"IlyaGusev/gemma-2-2b-it-abliterated is an 'abliterated' version of Google's Gemma-2-2B-IT model, created by Ilya Gusev. This model applies a specific methodology to modify the original Gemma-2-2B-IT, similar to the approach used for its 9B counterpart. It is designed to explore the effects of 'abliteration' on instruction-tuned language models, focusing on specific internal layers. The primary use case is research into model modification techniques and their impact on output generation.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":2600000000,"context_length":8192,"created_at":"2025-07-05T03:53:35.296Z","updated_at":"2026-10-15T01:23:23.947Z","hf_created_at":"2024-07-31T18:12:00.000Z","trending_rank":{"fetched_at":"2026-10-02T06:17:00.020Z","position":4084},"classified_tags":{"languages":["en"],"tasks":["text-generation","conversational"],"architectures":["gemma2"],"training":[],"capabilities":[],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"gemma2-2b","created_at":"2025-06-12T09:08:29.994Z","updated_at":"2026-01-27T07:35:50.732Z","concurrency_cost":1,"size":2.6,"parameter_size":2600000000,"context_length":8192,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma2","family_name":"Gemma 2"},"supports_tool_calling":false},{"id":"3MPER0RR/Qwen2.5-Coder-32B-Instruct-3MPER0RR-abliterated","name":"3MPER0RR/Qwen2.5-Coder-32B-Instruct-3MPER0RR-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a9fd98b8dfe2b89cd6d2d7f","favorites":0,"downloads":447,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The 3MPER0RR/Qwen2.5-Coder-32B-Instruct-3MPER0RR-abliterated model is a 32.8 billion parameter instruction-tuned causal language model based on the Qwen 2.5 Coder architecture, developed by 3MPER0RR. It is designed for coding-related tasks, leveraging its substantial parameter count and a 32K context length for complex programming challenges. This model is specifically optimized for instruction following in coding contexts, making it suitable for code generation, debugging, and understanding.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"2","parameter_size":32799999999.999996,"context_length":32768,"created_at":"2026-09-11T02:07:38.699Z","updated_at":"2026-10-14T23:33:18.348Z","hf_created_at":"2026-09-08T09:46:51.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen2.5-coder-32b-instruct","finetune:qwen/qwen2.5-coder-32b-instruct"]},"model_class":{"id":"qwen25-32b","created_at":"2025-06-12T09:08:32.600Z","updated_at":"2026-01-27T07:58:09.463Z","concurrency_cost":2,"size":32.8,"parameter_size":32799999999.999996,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"Momix-44/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated","name":"Momix-44/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69efe18fb3b43c2b62ffcc8c","favorites":2,"downloads":425,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The Momix-44/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated model is a 35.1 billion parameter language model based on the Qwen 3.6 architecture, specifically an uncensored version of lordx64/Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled. Developed by huihui-ai, this model has undergone \"abliteration\" to significantly reduce safety filtering and refusals, making it suitable for research into uncensored LLM behavior. It is quantized with an imatrix derived from the Bartowski Dataset and is primarily intended for experimental use in controlled environments where sensitive or controversial outputs are acceptable.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":35100000000,"context_length":32768,"created_at":"2026-06-15T22:12:45.511Z","updated_at":"2026-10-14T22:38:31.484Z","hf_created_at":"2026-04-27T22:22:07.000Z","trending_rank":null,"classified_tags":{"languages":["en"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen"],"training":["distillation","lora","unsloth","abliterated","uncensored"],"capabilities":["reasoning","chain-of-thought","moe","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_5_moe","qwen3.6","mixture-of-experts"],"base_models":["lordx64/qwen3.6-35b-a3b-claude-4.7-opus-reasoning-distilled","adapter:lordx64/qwen3.6-35b-a3b-claude-4.7-opus-reasoning-distilled"]},"model_class":{"id":"qwen3_5moe-35b","created_at":"2026-06-13T11:08:45.209Z","updated_at":"2026-06-13T11:08:45.209Z","concurrency_cost":2,"size":35.1,"parameter_size":35100000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3-5-moe","family_name":"Qwen3 5 Moe"},"supports_tool_calling":true},{"id":"IlyaGusev/gemma-2-9b-it-abliterated","name":"IlyaGusev/gemma-2-9b-it-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6692d59a4f911aa1afc37038","favorites":51,"downloads":403,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"gemma","categories":null,"tags":["niche"],"acc_tags":[],"description":"IlyaGusev/gemma-2-9b-it-abliterated is a 9 billion parameter instruction-tuned causal language model based on Google's Gemma 2 architecture. This model has been \"abliterated\" using a script based on TransformerLens, resulting in an uncensored version. It is primarily designed for use cases requiring a less restricted response generation, as demonstrated by its outputs to controversial queries.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":9000000000,"context_length":16384,"created_at":"2025-07-05T03:46:19.150Z","updated_at":"2026-10-15T01:23:12.294Z","hf_created_at":"2024-07-13T19:29:30.000Z","trending_rank":{"fetched_at":"2026-08-10T01:17:00.189Z","position":3006},"classified_tags":{"languages":["en"],"tasks":["text-generation","conversational"],"architectures":["gemma2"],"training":[],"capabilities":[],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"gemma2-9b","created_at":"2025-05-12T23:44:18.135Z","updated_at":"2026-03-09T13:45:02.013Z","concurrency_cost":1,"size":9,"parameter_size":9000000000,"context_length":16384,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":null,"family_name":null},"supports_tool_calling":false},{"id":"roslein/Qwen3-32B-abliterated","name":"roslein/Qwen3-32B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"681504fafb17bf4a9b06a6e5","favorites":13,"downloads":396,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The roslein/Qwen3-32B-abliterated model is a 32.8 billion parameter causal language model based on the Qwen3 architecture, featuring 64 layers and GQA. This version has undergone an \"abliteration\" process using proportional scaling to reduce refusal behavior while preserving some safety guardrails. It offers a balance between responsiveness and safety, making it suitable for research and applications where slight quality degradation is acceptable for broader response generation.","is_gated":false,"is_private":false,"is_conversational":false,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":true,"loadingStage":null,"sortRank":90,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":32000000000,"context_length":32768,"created_at":"2025-05-13T02:27:00.923Z","updated_at":"2026-10-15T01:01:38.075Z","hf_created_at":"2025-05-02T17:46:34.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":[],"architectures":["qwen3"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-32b","finetune:qwen/qwen3-32b"]},"model_class":{"id":"qwen3-32b","created_at":"2025-05-12T23:46:02.728Z","updated_at":"2026-01-27T07:29:53.291Z","concurrency_cost":2,"size":32,"parameter_size":32000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"shutkit/Qwen2.5-VL-7B-NSFW-Caption-V3-abliterated","name":"shutkit/Qwen2.5-VL-7B-NSFW-Caption-V3-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"68cde7b5217b07616822ff4c","favorites":7,"downloads":388,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The shutkit/Qwen2.5-VL-7B-NSFW-Caption-V3-abliterated is a 7 billion parameter vision-language model, derived from the Qwen2.5-VL family, with a 32768 token context length. This model has been specifically modified using abliteration techniques to remove refusal behaviors in its text generation, making it uncensored. It is primarily designed for generating captions for NSFW content without typical AI safety restrictions, focusing on the textual output rather than image processing.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":7000000000,"context_length":32768,"created_at":"2026-05-08T00:37:07.202Z","updated_at":"2026-10-14T22:22:41.534Z","hf_created_at":"2025-09-19T23:31:01.000Z","trending_rank":{"fetched_at":"2026-09-18T23:17:00.023Z","position":9905},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":[],"training":["uncensored","abliterated"],"capabilities":[],"modalities":["vision"],"domains":[],"creative":[],"content_flags":["not-for-all-audiences"],"other":["qwen2_5_vl"]},"model_class":{"id":"qwen25vl-7b","created_at":"2026-05-07T20:00:40.314Z","updated_at":"2026-05-07T20:00:40.314Z","concurrency_cost":1,"size":7,"parameter_size":7000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":false},{"id":"Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2","name":"Goekdeniz-Guelmez/Josiefied-Qwen2.5-7B-Instruct-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"66edd995e4b4a1c1cfc156ed","favorites":17,"downloads":387,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The Josiefied-Qwen2.5-7B-Instruct-abliterated-v2 is a 7.6 billion parameter instruction-tuned causal language model developed by Gökdeniz Gülmez, based on the Qwen2.5 architecture. This model is further fine-tuned on a custom dataset to enhance uncensored responses and instruction following, supporting a context length of up to 131,072 tokens. It is designed to act as a highly intelligent, capable, and fully uncensored AI assistant, optimized for productivity across various tasks.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":7600000000,"context_length":32768,"created_at":"2025-02-15T15:02:14.337Z","updated_at":"2026-10-15T00:58:32.660Z","hf_created_at":"2024-09-20T20:22:45.000Z","trending_rank":{"fetched_at":"2026-05-06T22:00:00.135Z","position":915407},"classified_tags":{"languages":["zh","en","fr","es","pt","de","it","ru","ja","ko","vi","th","ar"],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":[],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"arxiv":["2309.00071","2407.10671"],"base_models":["qwen/qwen2.5-7b","finetune:qwen/qwen2.5-7b"]},"model_class":{"id":"qwen25-7b","created_at":"2025-06-12T09:08:32.442Z","updated_at":"2026-03-09T13:58:41.753Z","concurrency_cost":1,"size":7.6,"parameter_size":7600000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"its-arrrpit/qwen-abliterated","name":"its-arrrpit/qwen-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69c050f99e9ee14d33688900","favorites":0,"downloads":382,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":null,"is_gated":false,"is_private":false,"is_conversational":false,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"2","parameter_size":7600000000,"context_length":32768,"created_at":"2026-03-24T01:47:59.002Z","updated_at":"2026-10-14T10:07:04.667Z","hf_created_at":"2026-03-22T20:28:41.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":[],"architectures":["qwen2"],"training":[],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[]},"model_class":{"id":"qwen2-7b","created_at":"2025-06-12T09:08:31.903Z","updated_at":"2026-03-09T13:58:05.398Z","concurrency_cost":1,"size":7.6,"parameter_size":7600000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen2","family_name":"Qwen 2"},"supports_tool_calling":true},{"id":"JunLee88999/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","name":"JunLee88999/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","owned_by":"Feather","root_model_id":null,"huggingface_id":"69a33a0892dbdec8ac87119f","favorites":1,"downloads":349,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"JunLee88999/Qwen3-4b-Z-Image-Turbo-AbliteratedV1 is a 4 billion parameter text encoder based on the Tongyi-MAI/Z-Image-Turbo architecture, fine-tuned to significantly reduce refusal rates for image generation and general prompts. Utilizing a 'p-e-w heretic method' over 1000 trials, it achieves a minimal KL Divergence of 0.0004 and a refusal rate of only 4% in torture tests. This model is specifically optimized for generating desired content without censorship, making it suitable for applications requiring unrestricted creative output.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"3","parameter_size":4000000000,"context_length":32768,"created_at":"2026-03-01T13:55:14.808Z","updated_at":"2026-10-14T21:49:49.589Z","hf_created_at":"2026-02-28T18:55:04.000Z","trending_rank":{"fetched_at":"2026-07-16T14:17:00.022Z","position":4570},"classified_tags":{"languages":[],"tasks":["conversational"],"architectures":["qwen3","qwen"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic","z-image","censorship-removal"],"base_models":["tongyi-mai/z-image-turbo","quantized:tongyi-mai/z-image-turbo"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"Goekdeniz-Guelmez/Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v2","name":"Goekdeniz-Guelmez/Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"68977a42d166a916cc1f13ae","favorites":24,"downloads":347,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"The Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v2 is a 4 billion parameter instruction-tuned language model developed by Goekdeniz-Guelmez, based on Alibaba's Qwen3 architecture. This model is part of the JOSIEFIED family, which utilizes a novel \"gabliteration\" technique to maximize uncensored behavior and instruction-following abilities. It is specifically fine-tuned for unrestricted, high-performance language generation, often outperforming its base counterparts on standard benchmarks.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":4000000000,"context_length":32768,"created_at":"2025-08-11T18:00:32.303Z","updated_at":"2026-10-15T01:27:34.646Z","hf_created_at":"2025-08-09T16:41:38.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":7852},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":[],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-4b-instruct-2507","finetune:qwen/qwen3-4b-instruct-2507"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"hotdogs/Qwen3.8-27B-thinkingcap-abliterated","name":"hotdogs/Qwen3.8-27B-thinkingcap-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a8c55afd0157b7e945074a2","favorites":5,"downloads":334,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"hotdogs/Qwen3.8-27B-thinkingcap-abliterated is a 27 billion parameter Qwen3.8-27B model, fine-tuned by hotdogs using a LoRA merge from the ThinkingCap efficient-reasoning dataset. This model is specifically designed to prevent 'runaway' thinking on complex, multi-step problems, significantly reducing output length without sacrificing accuracy. It excels at efficient reasoning by acting as 'runaway insurance' for hard problems, making it suitable for applications requiring concise and accurate responses.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-25T20:41:43.099Z","updated_at":"2026-10-15T00:31:21.107Z","hf_created_at":"2026-08-24T14:31:11.000Z","trending_rank":{"fetched_at":"2026-09-23T08:17:00.023Z","position":8120},"classified_tags":{"languages":[],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen3_5","qwen3"],"training":["lora","sft","abliterated"],"capabilities":["reasoning","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["thinkingcap","efficient-thinking","token-efficient","multi-token-prediction"],"datasets":["hotdogs/thinkingcap-sft-qwen38-27b"],"base_models":["hotdogs/qwen3.8-27b-abliterated","adapter:hotdogs/qwen3.8-27b-abliterated"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"mlabonne/gemma-3-12b-it-abliterated","name":"mlabonne/gemma-3-12b-it-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"67d70e5aee9f6c22e274295f","favorites":33,"downloads":333,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"gemma","categories":null,"tags":["niche"],"acc_tags":[],"description":"The mlabonne/gemma-3-12b-it-abliterated model is an uncensored 12 billion parameter instruction-tuned variant of Google's Gemma 3, featuring a 32768-token context length. Developed by mlabonne, this model utilizes a novel \"abliteration\" technique to remove refusals while largely preserving its original capabilities. It is specifically designed for applications requiring less restrictive content generation, offering a high acceptance rate for diverse prompts.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":12000000000,"context_length":32768,"created_at":"2025-07-05T05:25:11.243Z","updated_at":"2026-10-15T01:27:12.187Z","hf_created_at":"2025-03-16T17:46:02.000Z","trending_rank":{"fetched_at":"2026-10-14T19:17:00.030Z","position":7507},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["gemma3"],"training":[],"capabilities":[],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["google/gemma-3-12b-it","finetune:google/gemma-3-12b-it"]},"model_class":{"id":"gemma3-12b","created_at":"2025-04-26T14:03:40.946Z","updated_at":"2026-01-27T08:03:30.653Z","concurrency_cost":1,"size":12,"parameter_size":12000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma3","family_name":"Gemma 3"},"supports_tool_calling":false},{"id":"shuijiaodawang/Qwen3.8-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS","name":"shuijiaodawang/Qwen3.8-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a93e85da8fc5a9a1d1d762c","favorites":0,"downloads":303,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The shuijiaodawang/Qwen3.8-27B-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS is a 27 billion parameter multimodal derivative of Qwen/Qwen3.8-27B, developed by shuijiaodawang. This model is specifically engineered to significantly reduce refusal behavior across a wide range of prompts while maintaining its hybrid text-and-vision capabilities. It excels in usability, demonstrating 0 refusals across extensive internal prompt screens, making it suitable for applications requiring highly compliant and versatile AI responses.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"2","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-31T14:13:01.159Z","updated_at":"2026-10-14T23:27:41.275Z","hf_created_at":"2026-08-30T08:22:53.000Z","trending_rank":null,"classified_tags":{"languages":["multilingual"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":["qwen3_5","qwen"],"training":["abliterated","uncensored"],"capabilities":["multimodal","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3.8","refusal-reduction"],"base_models":["qwen/qwen3.8-27b","quantized:qwen/qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/Qwen3-4B-abliterated","name":"huihui-ai/Qwen3-4B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6811eda01dc84a443707e2e6","favorites":59,"downloads":291,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Qwen3-4B-abliterated model is an uncensored version of the Qwen/Qwen3-4B architecture, developed by huihui-ai. This 4-billion parameter model is specifically designed to remove refusal behaviors from the base LLM using an abliteration technique. It achieves a 100% pass rate on harmful instruction tests, making it suitable for applications requiring a less restrictive language model.","is_gated":true,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":4000000000,"context_length":32768,"created_at":"2025-07-05T05:05:13.453Z","updated_at":"2026-10-15T01:26:39.187Z","hf_created_at":"2025-04-30T09:30:08.000Z","trending_rank":{"fetched_at":"2026-10-12T08:17:00.028Z","position":5143},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-4b","finetune:qwen/qwen3-4b"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"fairy322/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","name":"fairy322/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a279f4bc314f6af639de058","favorites":1,"downloads":283,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The fairy322/Qwen3-4b-Z-Image-Turbo-AbliteratedV1 is a 4 billion parameter Qwen3-based text encoder, derived from Tongyi-MAI/Z-Image-Turbo. It has been specifically modified using a 'heretic method' to significantly reduce image generation and general refusal rates, achieving a 0.0004 KL Divergence and 4/100 refusal rate in torture tests. This model is optimized for generating desired content without censorship, making it suitable for applications requiring unrestricted output.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"3","parameter_size":4000000000,"context_length":32768,"created_at":"2026-06-12T16:09:50.823Z","updated_at":"2026-10-14T22:37:37.232Z","hf_created_at":"2026-06-09T05:06:19.000Z","trending_rank":{"fetched_at":"2026-07-16T14:17:00.022Z","position":5767},"classified_tags":{"languages":[],"tasks":["conversational"],"architectures":["qwen3","qwen"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic","z-image","censorship-removal"],"base_models":["tongyi-mai/z-image-turbo","quantized:tongyi-mai/z-image-turbo"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Rombo-TIES-v1.0","name":"BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Rombo-TIES-v1.0","owned_by":"Feather","root_model_id":null,"huggingface_id":"673e26b2bf93bea9ca8cd0fa","favorites":3,"downloads":269,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Rombo-TIES-v1.0 is a 32.8 billion parameter language model created by BenevolenceMessiah, merged using the TIES method. It is based on huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterated and incorporates rombodawg/Rombos-Coder-V2.5-Qwen-32b, which features self-instruct fine-tuning. This model is specifically designed for code-related tasks, leveraging its merged architecture to enhance coding capabilities.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"3","parameter_size":32799999999.999996,"context_length":32768,"created_at":"2026-01-25T13:28:10.882Z","updated_at":"2026-10-14T21:37:27.590Z","hf_created_at":"2024-11-20T18:13:06.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["mergekit","merge"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"arxiv":["2306.01708"],"base_models":["huihui-ai/qwen2.5-coder-32b-instruct-abliterated","merge:huihui-ai/qwen2.5-coder-32b-instruct-abliterated","rombodawg/rombos-coder-v2.5-qwen-32b","merge:rombodawg/rombos-coder-v2.5-qwen-32b"]},"model_class":{"id":"qwen25-32b","created_at":"2025-06-12T09:08:32.600Z","updated_at":"2026-01-27T07:58:09.463Z","concurrency_cost":2,"size":32.8,"parameter_size":32799999999.999996,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen25","family_name":"Qwen 2.5"},"supports_tool_calling":true},{"id":"wangzhang/Qwen3.8-27B-abliterated","name":"wangzhang/Qwen3.8-27B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a800457421f81c0de760620","favorites":12,"downloads":265,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"wangzhang/Qwen3.8-27B-abliterated is a 27 billion parameter Qwen3.8-27B derivative model, developed by Wangzhang Wu, specifically engineered to reduce refusal behavior. This model was created using a 10-round iterative LoRA self-distillation pipeline, rather than classical directional ablation, to teach it to imitate compliant completions. It is primarily intended for safety research, red-teaming, and robustness evaluation of LLMs by significantly lowering refusal rates.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":27000000000,"context_length":32768,"created_at":"2026-08-16T09:44:11.859Z","updated_at":"2026-10-15T01:04:56.472Z","hf_created_at":"2026-08-15T06:16:55.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":10316},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3_5_text","abliterix","lora-self-distillation"],"base_models":["qwen/qwen3.8-27b","finetune:qwen/qwen3.8-27b"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated","name":"huihui-ai/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69b558214e42dea6fad92060","favorites":129,"downloads":261,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliterated is a 27 billion parameter language model derived from Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled. This model has undergone an \"abliteration\" process to remove refusal behaviors, making it an uncensored variant. It is primarily intended for research and experimental use in environments where content filtering is not desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"5","parameter_size":27000000000,"context_length":32768,"created_at":"2026-06-15T21:23:06.611Z","updated_at":"2026-10-15T01:17:23.994Z","hf_created_at":"2026-03-14T12:44:17.000Z","trending_rank":{"fetched_at":"2026-10-13T05:17:00.030Z","position":8559},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["reasoning","chain-of-thought","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["claude","dense"],"base_models":["jackrong/qwen3.5-27b-claude-4.6-opus-reasoning-distilled","finetune:jackrong/qwen3.5-27b-claude-4.6-opus-reasoning-distilled"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"croll83/Qwopus3.6-27B-v1-Abliterated-preview","name":"croll83/Qwopus3.6-27B-v1-Abliterated-preview","owned_by":"Feather","root_model_id":null,"huggingface_id":"69ec11fb56f14d943df61da5","favorites":2,"downloads":253,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The croll83/Qwopus3.6-27B-v1-Abliterated-preview is an uncensored 27 billion parameter multimodal language model based on the Qwen3.6-27B architecture, featuring a 262K token context length and a vision encoder. This model is a Claude-distilled reasoning fine-tune of the original Qwen model, with its refusal direction removed using orthogonal-projection abliteration. It is designed for research and experimental use, excelling in structured reasoning and consistent answer styles across various tasks, including agentic coding and multimodal understanding.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"3","parameter_size":27000000000,"context_length":32768,"created_at":"2026-06-16T05:54:00.555Z","updated_at":"2026-10-15T01:31:56.193Z","hf_created_at":"2026-04-25T00:59:39.000Z","trending_rank":null,"classified_tags":{"languages":["en","it","zh"],"tasks":["conversational","image-text-to-text"],"architectures":["qwen3_5","qwen"],"training":["abliterated","uncensored"],"capabilities":["reasoning","instruct","vision-language-model","multimodal","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["qwen3.6","qwopus","turboquant","nvfp4","dflash"],"base_models":["jackrong/qwopus3.6-27b-v1-preview","quantized:jackrong/qwopus3.6-27b-v1-preview"]},"model_class":{"id":"qwen3_5-27b","created_at":"2026-06-12T11:29:33.224Z","updated_at":"2026-06-12T11:29:33.224Z","concurrency_cost":2,"size":27,"parameter_size":27000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true},{"id":"DuoNeural/Gemma-4-26B-A4B-Abliterated","name":"DuoNeural/Gemma-4-26B-A4B-Abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69d7627b858abba9d6d0ba40","favorites":3,"downloads":249,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"gemma","categories":null,"tags":["niche"],"acc_tags":[],"description":"DuoNeural/Gemma-4-26B-A4B-Abliterated is a 26 billion parameter Mixture-of-Experts (MoE) model, based on Google's Gemma-4-26B-A4B-it, with a 32768 token context length. Developed by DuoNeural, this model has undergone \"Abliteration\" via representation engineering to remove refusal behaviors related to content policy. It retains full reasoning, tool-use, and multilingual capabilities, making it suitable for applications requiring unconstrained instruction following.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"3","parameter_size":26000000000,"context_length":32768,"created_at":"2026-06-16T09:43:19.952Z","updated_at":"2026-10-14T22:39:22.149Z","hf_created_at":"2026-04-09T08:25:31.000Z","trending_rank":{"fetched_at":"2026-09-12T23:17:00.038Z","position":5369},"classified_tags":{"languages":["en","multilingual"],"tasks":["image-text-to-text","text-generation","conversational"],"architectures":[],"training":["abliterated","uncensored"],"capabilities":["moe","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["gemma4"],"base_models":["google/gemma-4-26b-a4b-it","finetune:google/gemma-4-26b-a4b-it"]},"model_class":{"id":"gemma4-25b","created_at":"2026-06-11T17:32:19.005Z","updated_at":"2026-06-11T17:32:19.005Z","concurrency_cost":2,"size":26,"parameter_size":26000000000,"context_length":32768,"is_moe":true,"active_parameters":4,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"gemma4","family_name":"Gemma 4"},"supports_tool_calling":true},{"id":"huihui-ai/Qwen3-14B-abliterated","name":"huihui-ai/Qwen3-14B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6812665f0bcbcbeef6969b7d","favorites":51,"downloads":244,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Qwen3-14B-abliterated model is a 14 billion parameter uncensored variant of the Qwen/Qwen3-14B large language model, developed by huihui-ai. This model has been modified using an 'abliteration' technique to remove refusal behaviors, achieving a 100% pass rate on harmful instruction tests. It is designed for applications requiring an LLM that does not refuse potentially harmful prompts, offering a 32768 token context length.","is_gated":true,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":14000000000,"context_length":32768,"created_at":"2025-07-05T04:32:43.888Z","updated_at":"2026-10-15T01:24:20.639Z","hf_created_at":"2025-04-30T18:05:19.000Z","trending_rank":{"fetched_at":"2026-10-14T06:17:00.030Z","position":7619},"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-14b","finetune:qwen/qwen3-14b"]},"model_class":{"id":"qwen3-14b","created_at":"2025-05-12T23:45:30.387Z","updated_at":"2026-03-01T08:03:00.238Z","concurrency_cost":1,"size":14,"parameter_size":14000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/Huihui-Qwen3-0.6B-abliterated-v2","name":"huihui-ai/Huihui-Qwen3-0.6B-abliterated-v2","owned_by":"Feather","root_model_id":null,"huggingface_id":"6853c9d883f4fb3371f5e3f7","favorites":7,"downloads":242,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3-0.6B-abliterated-v2 is an 0.8 billion parameter causal language model based on the Qwen3 architecture, developed by huihui-ai. This model is specifically designed as an uncensored version of Qwen/Qwen3-0.6B, created using an 'abliteration' method to remove refusal behaviors. It is optimized for research and experimental use cases where reduced safety filtering is desired, offering a proof-of-concept for uncensored LLM deployment.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":800000000,"context_length":32768,"created_at":"2025-12-17T23:40:24.461Z","updated_at":"2026-10-14T21:22:59.107Z","hf_created_at":"2025-06-19T08:27:04.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen3"],"training":["abliterated","uncensored"],"capabilities":["chat","tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["qwen/qwen3-0.6b","finetune:qwen/qwen3-0.6b"]},"model_class":{"id":"qwen3-0b6","created_at":"2025-06-12T09:08:32.830Z","updated_at":"2026-03-10T07:32:21.140Z","concurrency_cost":1,"size":0.8,"parameter_size":800000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated","name":"huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"6790c49f29aa9d3756434fd7","favorites":35,"downloads":234,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":null,"categories":null,"tags":["niche"],"acc_tags":[],"description":"huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated is an uncensored version of deepseek-ai/DeepSeek-R1-Distill-Qwen-14B, created by huihui-ai. This model utilizes an abliteration technique to remove refusal behaviors from the original DeepSeek-R1-Distill-Qwen-14B, serving as a proof-of-concept for uncensoring LLMs without TransformerLens. Its primary differentiator is the removal of refusals, making it suitable for use cases requiring less restrictive content generation. A newer version, huihui-ai/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2, is also available.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"4","parameter_size":14000000000,"context_length":32768,"created_at":"2025-02-15T12:10:23.442Z","updated_at":"2026-10-15T00:58:08.696Z","hf_created_at":"2025-01-22T10:12:47.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["text-generation","conversational"],"architectures":["qwen2"],"training":["abliterated","uncensored"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["deepseek-ai/deepseek-r1-distill-qwen-14b","finetune:deepseek-ai/deepseek-r1-distill-qwen-14b"]},"model_class":{"id":"qwen2-14b-lc","created_at":"2025-02-15T03:15:00.000Z","updated_at":"2025-02-15T03:15:00.000Z","concurrency_cost":1,"size":14,"parameter_size":14000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen2","family_name":"Qwen 2"},"supports_tool_calling":true},{"id":"fwwrsd/bennydaball-z-image-turbo-abliterated-v1","name":"fwwrsd/bennydaball-z-image-turbo-abliterated-v1","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a2c24351ef32c7359e679ba","favorites":0,"downloads":234,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The fwwrsd/bennydaball-z-image-turbo-abliterated-v1 is a 4 billion parameter instruction-tuned causal language model, based on the Tongyi-MAI/Z-Image-Turbo architecture, developed by BennyDaBall. This model is specifically modified to significantly reduce refusal rates for both image generation prompts and general queries, achieving a refusal rate of 4/100 in torture tests. It maintains a low KL Divergence of 0.0004, indicating minimal impact on its original capabilities while enhancing its ability to generate desired content without censorship. Its primary use case is for applications requiring less restrictive content generation, particularly for image-related tasks.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"2","parameter_size":4000000000,"context_length":32768,"created_at":"2026-06-15T17:34:12.331Z","updated_at":"2026-10-14T22:38:14.418Z","hf_created_at":"2026-06-12T15:22:29.000Z","trending_rank":null,"classified_tags":{"languages":[],"tasks":["conversational"],"architectures":["qwen3","qwen"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic","z-image","censorship-removal"],"base_models":["tongyi-mai/z-image-turbo","quantized:tongyi-mai/z-image-turbo"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"theworstgam3r/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","name":"theworstgam3r/Qwen3-4b-Z-Image-Turbo-AbliteratedV1","owned_by":"Feather","root_model_id":null,"huggingface_id":"6a360af2f830b32c6c82d6a4","favorites":1,"downloads":233,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The theworstgam3r/Qwen3-4b-Z-Image-Turbo-AbliteratedV1 is a 4 billion parameter text encoder, based on the Tongyi-MAI/Z-Image-Turbo model, specifically modified to reduce refusal rates in image generation and general responses. Utilizing a p-e-w heretic method applied over 1000 trials, this model achieves a low refusal rate of 4/100 with minimal impact on original model integrity (KL Divergence of 0.0004). It is designed for users seeking a less restrictive model for generating diverse content, particularly for image prompts.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027585674},"popularity_level":"3","parameter_size":4000000000,"context_length":32768,"created_at":"2026-06-28T09:58:58.589Z","updated_at":"2026-10-14T22:47:10.145Z","hf_created_at":"2026-06-20T03:37:22.000Z","trending_rank":{"fetched_at":"2026-07-16T14:17:00.022Z","position":6007},"classified_tags":{"languages":[],"tasks":["conversational"],"architectures":["qwen3","qwen"],"training":["abliterated"],"capabilities":["tool-use","function-calling"],"modalities":["text"],"domains":[],"creative":[],"content_flags":[],"other":["heretic","z-image","censorship-removal"],"base_models":["tongyi-mai/z-image-turbo","quantized:tongyi-mai/z-image-turbo"]},"model_class":{"id":"qwen3-4b","created_at":"2025-06-12T09:08:32.996Z","updated_at":"2026-03-10T07:24:18.754Z","concurrency_cost":1,"size":4,"parameter_size":4000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":false,"input_modalities":["text"],"output_modalities":["text"],"family":"qwen3","family_name":"Qwen 3"},"supports_tool_calling":true},{"id":"huihui-ai/gemma-3-12b-it-abliterated","name":"huihui-ai/gemma-3-12b-it-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"67e0d8c17a34afcaffb48bd3","favorites":31,"downloads":225,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"gemma","categories":null,"tags":["niche"],"acc_tags":[],"description":"huihui-ai/gemma-3-12b-it-abliterated is an uncensored variant of Google's Gemma-3-12B-IT, created using an 'abliteration' technique to remove refusal behaviors. This multimodal model, supporting both text and image inputs, is designed to provide direct responses without safety-aligned refusals. It maintains the original Gemma-3 architecture and is suitable for applications requiring unfiltered AI interactions.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"warm","isHotLive":false,"isHotRecent":false,"loadingStage":null,"sortRank":80,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":12000000000,"context_length":32768,"created_at":"2025-06-28T03:36:06.823Z","updated_at":"2026-10-15T01:22:05.216Z","hf_created_at":"2025-03-24T04:00:01.000Z","trending_rank":{"fetched_at":"2026-10-08T06:17:00.029Z","position":4538},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["gemma3"],"training":["abliterated","uncensored"],"capabilities":[],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":[],"base_models":["google/gemma-3-12b-it","finetune:google/gemma-3-12b-it"]},"model_class":{"id":"gemma3-12b","created_at":"2025-04-26T14:03:40.946Z","updated_at":"2026-01-27T08:03:30.653Z","concurrency_cost":1,"size":12,"parameter_size":12000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text"],"output_modalities":["text"],"family":"gemma3","family_name":"Gemma 3"},"supports_tool_calling":false},{"id":"huihui-ai/Huihui-Qwen3.5-9B-Claude-4.6-Opus-abliterated","name":"huihui-ai/Huihui-Qwen3.5-9B-Claude-4.6-Opus-abliterated","owned_by":"Feather","root_model_id":null,"huggingface_id":"69be9a3a642a709caddd65cf","favorites":64,"downloads":217,"avg_rating":0,"total_reviews":0,"status":"active","health":null,"license":"apache-2.0","categories":null,"tags":["niche"],"acc_tags":[],"description":"The huihui-ai/Huihui-Qwen3.5-9B-Claude-4.6-Opus-abliterated is a 9 billion parameter language model derived from Jackrong's Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled-v2. This model has undergone an 'abliteration' process to significantly reduce its safety filtering and remove refusals, making it an uncensored variant. With a 32768-token context length, it is primarily intended for research and experimental use in controlled environments where unfiltered outputs are desired.","is_gated":false,"is_private":false,"is_conversational":true,"is_loaded":false,"availability":{"tier":"loading","isHotLive":false,"isHotRecent":false,"loadingStage":"init","sortRank":70,"lastChangedAt":1792027902332},"popularity_level":"4","parameter_size":9000000000,"context_length":32768,"created_at":"2026-06-15T23:13:57.240Z","updated_at":"2026-10-14T23:48:38.273Z","hf_created_at":"2026-03-21T13:16:42.000Z","trending_rank":{"fetched_at":"2026-10-15T01:17:00.129Z","position":8773},"classified_tags":{"languages":[],"tasks":["image-text-to-text","conversational"],"architectures":["qwen3_5"],"training":["abliterated","uncensored"],"capabilities":["reasoning","chain-of-thought","tool-use","function-calling"],"modalities":["vision"],"domains":[],"creative":[],"content_flags":[],"other":["claude","dense"],"base_models":["jackrong/qwen3.5-9b-claude-4.6-opus-reasoning-distilled-v2","finetune:jackrong/qwen3.5-9b-claude-4.6-opus-reasoning-distilled-v2"]},"model_class":{"id":"qwen3_5-9b","created_at":"2026-06-12T10:59:16.626Z","updated_at":"2026-06-12T10:59:16.626Z","concurrency_cost":1,"size":9,"parameter_size":9000000000,"context_length":32768,"is_moe":false,"active_parameters":null,"vision_supported":true,"input_modalities":["text","image"],"output_modalities":["text"],"family":"qwen3-5","family_name":"Qwen3 5"},"supports_tool_calling":true}],"pagination":{"current_page":1,"per_page":100,"total_items":558,"total_pages":6}}