Alibaba's Qwen3.8-Max-0902 Coding Update Tops Code Arena WebDev, Edging Out Claude Opus 5 Max

Alibaba released Qwen3.8-Max-0902 on September 2, a post-training snapshot upgrade to its flagship Qwen3.8-Max model focused specifically on coding and long-horizon agent tasks. The underlying architecture is unchanged — 2.4 trillion total parameters, a 1-million-token context window, and the same $2/$6 per-million-token API pricing — but Alibaba said all eight tracked coding benchmarks improved, with the largest gains on TerminalBench 3.0 (11.3 to 29.0) and ProgramBench Almost Solved (10.5 to 28.0), both more than doubling. On the Code Arena WebDev leaderboard, Qwen3.8-Max-0902 ranked first overall with 1,691 points, narrowly ahead of Anthropic's Claude Opus 5 Max (1,687), Moonshot's Kimi K3 Max (1,674), and its own predecessor Qwen3.8-Max (1,669). The release extends the rapid-iteration pattern among Chinese AI labs — Alibaba, DeepSeek, Moonshot, Zhipu, and Tencent have all shipped major open-weight model updates within the past two weeks as competition intensifies for share of the coding-agent market.