Qwen 3.8-27B
Released on August 14, 2026
Changes from previous version
A compact dense multimodal alternative to the 2.4T MoE Qwen3.8-2.4T-A95B, aimed at ordinary GPU hosts under Apache 2.0.
Release Summary
Dense 27B vision-language sibling of Qwen3.8-Max for local deployment. Apache 2.0 weights, 262K native context (extensible to 1M), image and video input, and 73.0 on Terminal-Bench 2.1 / 61.7 on SWE-bench Pro in Qwen's post-training evals.
Timeline
August 14, 2026
Qwen3.8-27B released
Qwen publishes the 27B dense multimodal checkpoint on Hugging Face and ModelScope under Apache 2.0.