Qwen3.7 Max leads the MMLU ranking

Qwen3.7 Max ranks first on the MMLU benchmark as of August 15, 2026, scoring 93.7%. GPT-5 follows at 93.5%, while o3 ranks third at 93.1%.

MMLU, or Massive Multitask Language Understanding, tests knowledge across 57 subjects and is categorized as a general knowledge benchmark. The ranking includes 77 models, with a best score of 93.7, an average score of 82.4, and a standard deviation of 12.7.

The dataset pairs benchmark results with provider pricing, including input and output costs per million tokens. Qwen3.7 Max is listed at $1.250 input and $3.750 output, while GPT-5 is listed at $1.250 input and $10.000 output.