Qwen 3.6 27B

• - 🆕 Dense, native vision-language model from Alibaba's Qwen3.6 family, 2026.
• - 👁️ Natively multimodal: accepts text, image, and video inputs.
• - 📏 256,000-token context window.
• - 🔧 Improved agentic coding over the prior 3.5-27B release.
• - 🧠 Enhanced STEM reasoning and inference capabilities.
• - 🎯 Advances in spatial intelligence, object localization, and detection.
• - 📚 Steady gains in video understanding and document OCR.
• - ⚡ Served in FP8 quantization; supports function calling and web search.

Alibaba

Alibaba Group is a Chinese multinational technology company founded in 1999 and headquartered in Hangzhou, Zhejiang. Originally built around e-commerce and cloud computing, Alibaba has become one of the most prolific contributors to open-weight AI research,…

Qwen 3.6 27B is a dense, natively multimodal vision-language model in Alibaba's Qwen3.6 family, accepting text, image, and video inputs. According to the catalog description, it builds directly on the earlier Qwen3.5-27B model, extending that lineage rather than introducing a new model class.

Compared with same-family predecessors such as Qwen 3.5 9B and the much larger Qwen 3.5 397B, the documented changes over the 3.5-27B version center on agentic coding and enhanced STEM reasoning and inference. In the vision modality, the catalog notes advances in spatial intelligence, object localization, and detection, with steady improvements to video understanding, document OCR, and visual agent capabilities.

The model offers a 256,000-token context window and is served in FP8 quantization for efficient inference. Its listed capabilities include reasoning, vision, code-optimized generation, function calling, and web search, making it suited to multimodal agent and tool-use workflows.

It sits alongside larger sparse siblings such as Qwen 3.6 35B A3B FP8 and the proprietary Qwen 3.7 Max within Alibaba's broader Qwen lineup.