Skip to content

EthenEthenEthen

LoginPlatformTry Ethen

Open Source Model Profile · huihui-ai

Huihui-Qwen3-VL-32B-Thinking-abliterated

Huihui-Qwen3-VL-32B-Thinking-abliterated is a 33.36B-parameter Qwen3-VL thinking derivative from huihui-ai with publisher-described refusal abliteration.

Publisher
huihui-ai
Task
image-text-to-text
Model type
qwen3_vl
License
apache-2.0
Library
transformers
Publication status
Accepted · not indexed

Model overview

Huihui-Qwen3-VL-32B-Thinking-abliterated is published by huihui-ai as a qwen3_vl image-text-to-text model. The captured configuration identifies Qwen3VLForConditionalGeneration and Safetensors metadata reports 33,357,390,064 parameters. According to the model card, it is an uncensored abliterated derivative of Qwen/Qwen3-VL-32B-Thinking with only the text part processed, and card data records apache-2.0.

Recorded capabilities

Abliterated thinking-model derivative

According to the model card, the release applies abliteration to Qwen/Qwen3-VL-32B-Thinking, processing only the text part and not the image part.

33B-class vision-text build

The captured configuration identifies Qwen3VLForConditionalGeneration with model type qwen3_vl and Safetensors metadata reports about 33.36B parameters.

Documented inference workflow

According to the model card, inference uses a processor chat template with tokenization and model.generate, including a max_new_tokens example.

Explicit publisher safety warnings

According to the model card, reduced safety filtering creates sensitive-output and legal-ethical risks, and users are solely responsible for consequences.

Use cases in the source record

  • Image-text-to-text generation experiments following the publisher's documented processor and chat-template workflow.
  • Controlled safety-behavior research under the publisher's warnings, with rigorous output review and legal compliance.

Limitations and unknowns

  • No evaluation results were extracted from this record as structured benchmarks.
  • No context-window value was extracted from this record.
  • Provider state is historical snapshot data and should be refreshed before being presented as current.
  • Leaderboard and refusal-removal claims come from the publisher model card and were not independently verified by Ethen.

Source and provenance

Source: huihui-ai/Huihui-Qwen3-VL-32B-Thinking-abliterated

Captured: Unknown. Processed: 2026-09-07T19:35:56.746649+00:00.

huihui-ai/Huihui-Qwen3-VL-32B-Thinking-abliterated This is an uncensored version of Qwen/Qwen3-VL-32B-Thinking created with abliteration (see remove-refusals-with-transformers to know more about it). It was only the text part that was processed, not the image part. The abliterated model will no longer say "I can’t describe or analyze this image." UGI Leaderboard This model now tops the UGI Leaderboard with a perfect score in the W10 category. Meaning that it's the first model to not refuse to answer or to deviate from an unsafe instruction at all. ollama Please update to the latest version of Ollama-v0.12.7 . You can use huihui_ai/q…

F001F002F003F004F005F006F007F009F010F011F012F013F015F016F017