A community-modified variant of Qwen's 27B model that has undergone abliteration — a technique that removes refusal behaviors from the base model. This means it responds more freely to a wider range of prompts without built-in content restrictions. It handles both text and image inputs, inheriting the multimodal capabilities of its Qwen 3 lineage.