Alibaba's Qwen3.8-Max Lands Arena #4, Alibaba Shares Pop 6% 🚀

Alibaba launched Qwen3.8-Max on Monday with its first published benchmarks, placing the 2.4 trillion-parameter model fourth on Arena's Frontend Code leaderboard with 1,668 points, trailing only Claude Opus 5 (Max) at 1,705 points and Kimi K3 (Max) at 1,676 points, and tying Claude Opus 5 (High) at 1,669 points. Qwen3.8-Max also secured second place in Arena's Consumer Product category, according to figures shared by Arena.ai on August 3, 2026.

The model exits its July preview with 95 billion active parameters and a 1 million-token context window, with Alibaba reporting scores of 86.6 on TerminalBench-2.1, 93.0 on PaperBench, and 86.1 on OSWorld-Verified, alongside claimed wins over Claude Opus 4.8, Fable 5, and GPT-5.6 on agentic and multimodal benchmarks. Anthropic's Fable 5 maintained a wide lead in core software engineering at 80.0 on SWE-Pro against 67.7 for Qwen3.8-Max, and 88.8 on FrontierSWE against 73.5.

API pricing debuted at $2 per million input tokens and $6 per million output tokens, and the launch also marks the first time Alibaba will open-source the weights of a Qwen-Max-class model. "Next week, the open weights of Qwen3.8-Max will be released, and Qwen3.8-27B is also going open-weights to meet you all!" the team said, confirming the weights will land on Hugging Face and ModelScope.

Investor reaction was swift, with Alibaba shares climbing 6.15% to HK$124.20 by the Hong Kong lunch break from a previous close of HK$117, extending Friday's 4.65% gain that analysts tied to a reported Moonshot chip deal. Independent evaluations are expected to sharpen the picture once the weights are released next week.