Huihui-Step3-VL-10B-abliterated-GGUF

This repository contains GGUF quantized versions of huihui-ai/Huihui-Step3-VL-10B-abliterated.

These files allow you to run the model locally on consumer hardware (CPU and Mac/PC GPUs) using llama.cpp or other compatible inference software.

⚠️ Note on Vision Capabilities

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."

🌟 About the Original Model

The original model is an uncensored (abliterated) version of stepfun-ai/Step3-VL-10B [1].

• Refusals Removed: The text component of the model was processed using the abliteration technique (orthogonalized against refusal directions). As a result, the model will no longer reply with standard refusals like "I can't describe or analyze this image." when prompted with potentially sensitive images [1].
• Capabilities: Multimodal / Vision-Language (Image-Text-to-Text) [1].
• Language: Multilingual / Conversational [1].

Note: Only the text-generation part was abliterated; the vision encoder retains its original mapping [1].

🚀 How to Run

You can run this model using llama.cpp or UI tools based on it (such as LM Studio, Ollama, text-generation-webui, etc.).

Using llama.cpp (Command Line)

Because this is a Vision-Language Model (VLM), you must pass the multimodal projector file (--mmproj) alongside the main model to process images.

# Example command using llama-cli
./llama-cli \
  -m Huihui-Step3-VL-10B-abliterated-Q4_K_M.gguf \
  --mmproj Huihui-Step3-VL-10B-abliterated-mmproj-f16.gguf \
  --image path/to/your/image.png \
  -p "Describe this image in detail." \
  -c 4096 \
  -temp 0.7

Note: Please ensure you are using a recent version of llama.cpp that supports the Step3-VL architecture.

⚠️ Usage Warnings (From the Original Author)

• Risk of Sensitive or Controversial Outputs: This model's safety filtering has been significantly reduced, potentially generating sensitive, controversial, or inappropriate content [1]. Users should exercise caution and rigorously review generated outputs.
• Not Suitable for All Audiences: Due to limited content filtering, the model's outputs may be inappropriate for public settings, underage users, or applications requiring high security [1].
• Legal and Ethical Responsibilities: Users must ensure their usage complies with local laws and ethical standards. Generated content may carry legal or ethical risks, and users are solely responsible for any consequences [1].
• Research and Experimental Use: It is recommended to use this model for research, testing, or controlled environments, avoiding direct use in production or public-facing commercial applications [1].
• No Default Safety Guarantees: Unlike standard models, this model has not undergone rigorous safety optimization. The authors bear no responsibility for any consequences arising from its use [1].

💖 Credits

• Original Base Model: stepfun-ai [1]
• Abliteration processing: huihui-ai [1]
• GGUF Quantization: 2lains