Qwen 3.8 Max

Alibaba introduced Qwen 3.8 Max as its flagship artificial intelligence model, featuring 2.4 trillion parameters and a focus on coding, long-horizon agentic workflows, and multimodal reasoning. The company announced plans to release open weights for the model and a smaller 27-billion-parameter version. API pricing was set at $2.00 per million input tokens, $6.00 per million output tokens, and $0.25 per million cached tokens.

The model achieved strong placements on several third-party leaderboards. It ranked fourth in the Frontend Code Arena with an Elo score of 1,668 and second in the Vision Arena. On the Vals Index, it scored 66.1, matching Claude Opus 4.7 while operating at a lower cost per test, and scored 87.3% on SWE-bench. Observers noted that its architecture uses sparse mixture-of-experts technology to activate a small fraction of its total parameters per token, helping reduce serving costs while supporting a 1-million-token context window.

Despite strong benchmark results, critics highlighted significant infrastructure challenges, noting that operating a model of this size requires extensive hardware resources such as multi-GPU clusters. Additional caution centered on licensing terms, with some users flagging geographic restrictions that could limit usage in major Western markets. Simultaneously, developers pointed out that the release of the smaller 27-billion-parameter variant may offer a more practical option for local deployment and broader ecosystem adoption.