Qwen3.7-Max-Preview

Alibaba's new flagship reasoning model, unveiled at the 2026 Alibaba Cloud Summit on May 20 after debuting anonymously on LM Arena around May 14. Proprietary, preview-only, API access via Alibaba Cloud; no open weights as of release. 1M token context with always-on extended-thinking reasoning. Successor to Qwen3.6-Plus in the closed-flagship slot.

AA Intelligence Index v4.0: 57 (#7 of 150 models, #1 among Chinese labs as of May 2026; up 5 points from Qwen3.6-Max-Preview's 52). LM Arena text Elo ~1475 (#13 overall), top-10 on math and coding subleaderboards. Pricing: $2.50 per 1M input / $7.50 per 1M output tokens. Verbose by design — 97M output tokens to score the AA index vs a 36M-token median, consistent with the lab's pitch of long-horizon agentic execution (claims up to 35-hour autonomous runs and 1000+ tool calls per task).

No technical report or arXiv ID at launch.