Qwen 3.7 Max Ships, Tops Intelligence Index Above Gemini 3.5 Flash
Alibaba released Qwen 3.7 Max, which landed in the top 10 of the artificial-analysis intelligence index composite, scoring above Gemini 3.5 Flash. In an independent test, the model solved Discover AI's proprietary 12-month-old elevator-puzzle benchmark — three interwoven optimization tracks — on first attempt, validated its own result, and found a shorter optimized solution on a second pass. Generation-over-generation jumps include the reasoning benchmark score rising from 8 (Qwen 3.6 Plus) to 44 (Qwen 3.7 Max). The model deliberately obfuscates reasoning traces to prevent chain-of-thought distillation.
Why It Matters
These benchmark jumps combined with real-world test success position Qwen 3.7 Max as a significant challenger in the frontier tier, further fragmenting a competitive landscape previously dominated by OpenAI and Anthropic. Alibaba's open-availability approach means the gains are immediately accessible to the developer community.