Qwen3.5-9B

This model was trained in BF16/TF32 with Adam on Chain of Thought ONLY This model shows a massive increase in uncensored and unfiltered results without any direct image to text training or assistance response training. In testing around 5 out of every 100 images would return with the censored result. Compared to 98 out of 100 returning censored on the untrained model.

• Default "Describe this image" is recommended - Putting NSFW info in the prompt is possible for context.

Note: This model is intended to be run at 512 tokens utilizing the THINK portion of the model NOT the assistant response. Exceeding 1024 tokens will likely diminish the results not enhance them.