AI BriefingPickQwen-Image-2.1-Turbo-Uncensored-GGUF: Qwen-Image Text Encoder: Reducing Refusal Rate from 88.9% to 1.2%
AtomicChat/Qwen-Image-2.1-Turbo-Uncensored-GGUF
About the project
This is a GGUF model that functions as the text encoder for the Qwen-Image pipeline in the stable-diffusion.cpp environment. By removing the refusal bias from the original Qwen3-VL-8B-Instruct, the refusal rate for English was reduced from 88.9% to 1.2%.
It is designed to bypass filtering while maintaining text generation capabilities. The MMLU score remains unchanged at 77.35%, and image generation results are guaranteed to be of the same quality as the original model.
It includes various quantization versions from Q2_K to BF16 and multimodal projector files. It can be integrated with Atomic Chat or llama.cpp to build an unrestricted image generation workflow locally.
AtomicChat/Qwen-Image-2.1-Turbo-Uncensored-GGUF
The original page has no description.
text-to-image
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.