license: apache-2.0
language:

• en
  pipeline_tag: text-generation
  tags:
• vision
• fp8
• qwen
• qwen3vl
• abliterated
• safetensors

Qwen3-VL 8B Abliterated FP8 Scaled

This repository contains the qwen3vl_8b_abliterated_fp8_scaled.safetensors model, a fine-tuned, abliterated, and FP8 scaled version of the Qwen3-VL 8B model. It is optimized for environments like ComfyUI, providing excellent performance with a reduced memory footprint.

Model Details

• Base Model: Qwen3-VL 8B
• Modifications:
  • Abliterated: Safety alignments and refusal mechanisms have been significantly reduced or removed.
  • FP8 Scaled: Quantized and scaled to FP8 (8-bit floating point) for lower VRAM usage and faster inference speed, while retaining high quality.
• Format: .safetensors

Usage in ComfyUI

To use this model in ComfyUI:

1. Place the qwen3vl_8b_abliterated_fp8_scaled.safetensors file in your ComfyUI/models/text_encoders/ or appropriate models directory depending on the nodes you use.
2. Load it using the appropriate text encoder or vision loader node.

Disclaimer

This model has been "abliterated," meaning standard safety alignments present in the original model may be bypassed. Please use responsibly and ensure your usage complies with applicable laws and regulations.