Qwen 3.6 27B

Qwen 3.6 27B is the dense flagship of Alibaba's April 2026 open-weight generation — a 27B vision-language model mixing Gated DeltaNet and Gated Attention layers, with 262K native context (extensible to ~1M via YaRN) and a togglable hybrid thinking mode. Its headline result is 77.2 on SWE-bench Verified, beating Alibaba's own far larger Qwen 3.5 397B MoE on agentic coding. It handles text, image, and video input and posts 94.1 on AIME 2026 and 87.8 on GPQA Diamond, making it arguably the strongest quality-per-GB model available. The Q4_K_M build fits a 24 GB GPU (RTX 3090/4090) or a 32 GB Mac; Q8_0 wants 34 GB and suits 48 GB Macs or dual-GPU rigs.