According to CoinMeta, Qianwen has officially open-sourced Qwen3.8-27B, and the official benchmark scores have also been released. In the 8 directly comparable tests listed by the official team, Qwen3.8-27B outperforms Meta's newly released 30B model Muse Glimmer in all aspects, with the gap being particularly evident in agent and programming. Qwen3.8-27B scored 73.0 in terminal-bench 2.1, while Muse Glimmer scored 51.7. In swe-bench and pro, Qwen3.8-27B scored 61.7 against 51.2. Qwen3.8-27B's independent benchmark score is 52 points, tying it with DeepSeek V4 Flash and GPT-5.6. It is only 1 point behind DeepSeek V4 Pro and GLM-5.2's score of 53 points. The official default inference setting for Qwen is xhigh, and it also supports medium, low, as well as the option to turn off thinking.