Skip to content

Agents' Last Exam

An agent benchmark reported in DeepSeek's V4 Flash 0731 launch comparison.

17 models

Higher is better.

  1. 1GPT-6 AstraOpenAI59.3
  2. 2GPT-6 SolOpenAI56.4
  3. 3Qwen3.8 MaxAlibaba52.4
  4. 4Qwen3.8-Flash-NextAlibaba51.2
  5. 5Qwen3.8-27BAlibaba42.9
  6. 6Gemini 4 ArgonGoogle39.5
  7. 7DeepSeek V4.1 Flash 1017DeepSeek35.4
  8. 8MiMo-V2.6-ProXiaomi31.6
  9. 9MiniMax M3.1 FlashMiniMax29.6
  10. 10Step 5 PreviewStepFun29.5
  11. 11GLM-5.3Z.AI28.5
  12. 12MiMo-V2.6-FlashXiaomi27.6
  13. 13Gemini 3.7 FlashGoogle26.3
  14. 14GLM-5.3-FlashZ.AI26.3
  15. 15DeepSeek V4 Pro 0813DeepSeek25.7
  16. 16DeepSeek V4 Flash 0731DeepSeek25.2
  17. 17Hy4 previewTencent22.8
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Provider-reported task score
Tasks
Agent tasks
Authors
DeepSeek-AI (2026)

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