A mid-sized multimodal model that accepts both text and image inputs, producing text output. It carries the 'abliterated' tag, meaning refusal behaviors have been removed from the base Qwen3 architecture, so it responds more openly to a wider range of prompts without built-in content restrictions. The NVFP4A16 designation indicates a mixed-precision quantization format optimized for NVIDIA hardware, trading some numerical precision for reduced memory footprint.