license: apache-2.0
language:

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

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

The Qwen3-VL-2B-Thinking-abliterated-v1 from prithivMLmods is a 2B-parameter vision-language model variant of Alibaba's Qwen3-VL-2B-Thinking, modified through abliteration (v1.0) to eliminate safety refusals and content filters, enabling uncensored, step-by-step visual reasoning and detailed captioning across complex, sensitive, artistic, technical, or abstract imagery while retaining the base model's advanced multimodal perception, instruction-following, and "thinking" chain-of-thought capabilities. Optimized for variational output control—from brief summaries to intricate, multi-level analyses—it robustly handles diverse resolutions, aspect ratios, and layouts, primarily in English with prompt-based multilingual support, making it ideal for research, red-teaming, creative visual generation, and unrestricted agentic applications. This abliterated version delivers high-fidelity, factual descriptions without conventional guardrails, supporting efficient local inference for advanced visual reasoning tasks.

Qwen3-VL-2B-Thinking-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):