license: apache-2.0
base_model:

• prithivMLmods/Qwen3-VL-2B-Instruct-abliterated-v1
  language:
• en
  pipeline_tag: image-text-to-text
  library_name: transformers
  tags:
• text-generation-inference
• llama.cpp
• abliterated

Qwen3-VL-2B-Instruct-abliterated-v1-GGUF

The Qwen3-VL-2B-Instruct-abliterated-v1 from prithivMLmods is a 2B-parameter vision-language model variant of Alibaba's Qwen3-VL-2B-Instruct, modified through abliteration (v1.0) to remove safety refusals and content filters, enabling uncensored, detailed captioning and multimodal reasoning across sensitive, complex, artistic, technical, or abstract visual content while preserving the base model's strong instruction-following and visual agent capabilities. Optimized for high-fidelity descriptions with variational detail control—from concise summaries to intricate, step-by-step narratives—it handles diverse aspect ratios, resolutions, and layouts robustly, supporting primarily English outputs with multilingual adaptability via prompt engineering for research, red-teaming, and creative applications. Built on Qwen3-VL's advanced image comprehension, this abliterated version bypasses conventional guardrails to deliver factual, descriptive responses without censorship, making it suitable for unrestricted visual analysis in local deployments.

Qwen3-VL-2B-Instruct-abliterated-v1 [GGUF]

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant
types (lower is better):