Qwen3.7-Max Hits #4 on Code Arena, On Par With Claude Opus 4.6

Alibaba's Qwen team announced on May 27 that Qwen3.7-Max debuted at #4 on Code Arena with a score on par with Claude Opus 4.6, making it the top-ranked Chinese lab on the board. The team teased 'more to ship' in coming weeks.

The placement is notable because Code Arena has become the most-watched developer-preference leaderboard, and Chinese labs have historically punched below their weight there compared to benchmark-style evaluations. A debut at #4 — ahead of older Anthropic models, multiple Google checkpoints, and various open-source contenders — closes the gap meaningfully even if Opus 4.8 (released the next day) raises the bar again.

Competitive frame: the same week, xAI claimed Grok V9 finished training heavy on Cursor coding data, OpenAI's Greg Brockman called GPT-5.5 'uniquely good' for coding, and Anthropic shipped Opus 4.8 at 88.6% SWE-bench. Coding is the contested benchmark of the quarter, and Qwen is now demonstrably in that race.

Watch: whether Qwen3.7-Max appears in mainstream Western developer tooling (Cursor, Windsurf, Cline) defaults, and how the upcoming Qwen releases hinted at by the team compare to Opus 4.8.