Skip to content

Harvard-MIT Mathematics Tournament February 2026

A February 2026 HMMT slice used in newer frontier-model math comparisons.

24 models

Higher is better.

  1. 1Qwen3.7 MaxAlibaba97.1
  2. 2DeepSeek V4 Pro 0813DeepSeek95.2
  3. 3DeepSeek V4 Flash 0731DeepSeek94.8
  4. 4Solar Open 2Upstage93.9
  5. 5Qwen3.7 PlusAlibaba92.9
  6. 6Kimi K2.6Moonshot AI92.7
  7. 7GLM-5.2Z.AI92.5
  8. 8Inkling-SmallThinking Machines Lab90.2
  9. 9Qwen3.5 397BAlibaba87.9
  10. 10Qwen3.6 PlusAlibaba87.8
  11. 11Kimi K2.5Moonshot AI87.1
  12. 12Ling 3.0 FlashInclusionAI87.0
  13. 13GLM-5Z.AI86.4
  14. 14Claude Opus 4.5Anthropic85.3
  15. 15MAI-Thinking-1Microsoft84.9
  16. 16Qwen3.6-27BAlibaba84.3
  17. 17Qwen3.6-35B-A3BAlibaba83.6
  18. 18GLM-5.1Z.AI82.6
  19. 19K-EXAONE 2.0LG AI Research78.4
  20. 20MiniMax M2.7MiniMax72.7
  21. 21ZAYA1-8BZyphra71.6
  22. 22MiniCPM5-2BOpenBMB63.8
  23. 23LongCat-Flash-Lite-SparseMeituan40.5
  24. 24MiniCPM5-1BOpenBMB25.8
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Contest mathematics
Tasks
Competition math problems
Authors
Qwen (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.