A fine-tuned variant built for fewer content restrictions, this model carries the Qwen 3 family's multilingual reasoning and instruction-following capabilities while relaxing the guardrails typical of base releases. It handles text and image inputs, making it multimodal in practice. The 'uncensored' framing means it will engage with topics that standard instruct models decline, which shifts responsibility for appropriate use entirely to the deployer.