Alibaba Drops Qwen3.8-Max Weights Next Week, Wallets Its API Anyway

Alibaba launched Qwen3.8-Max on Monday, publishing first benchmarks and confirming that open weights for the model — along with Qwen3.8-27B — will be released "next week" on Hugging Face and ModelScope. It marks the first time Alibaba has open-sourced a Qwen-Max-class model. The system carries 2.4 trillion total parameters with 95 billion active, a 1 million-token context window, and an API priced at $2 per million input tokens and $6 per million output tokens.

On Arena's Frontend Code leaderboard, Qwen3.8-Max scored 1,668 points to place fourth. Claude Opus 5 (Max) led at 1,705, Kimi K3 (Max) held second at 1,676, and Claude Opus 5 (High) sat one point ahead of Qwen at 1,669. The model also took second place in Arena's Consumer Product category.

Alibaba shared internal data claiming wins over Claude Opus 4.8, Fable 5, and GPT-5.6 on agentic and multimodal benchmarks, posting 86.6 on TerminalBench-2.1, 93.0 on PaperBench, and 86.1 on OSWorld-Verified. On core software engineering, Anthropic's Fable 5 led with 80.0 on SWE-Pro versus Qwen's 67.7, and 88.8 on FrontierSWE against 73.5. Alibaba also reported the model spent 16 days building a coding tool autonomously — 265 commits, 127 pull requests, and 151 issues — and finished ahead of 458 of 526 human teams in a 24-hour machine learning contest.

Investors reacted to the launch. Alibaba shares rose 6.15% to HK$124.20 by the Hong Kong lunch break, up from a previous close of HK$117 and extending Friday's 4.65% gain that analysts linked to a reported Moonshot chip deal.