Wikibench

IFBench

general

Instruction Following Benchmark evaluating model's ability to follow complex instructions

Leaderboard

Showing 20 of 24 results

  1. Hermes 3 70B
    81.2%
    i
  2. Nova 2 Pro
    80.2%
    i
  3. Qwen3.7 Max
    79.1%
    i
  4. Qwen3.5-27B
    76.5%
    i
  5. Qwen3.5-397B-A17B
    76.5%
    i
  6. Qwen3.5-122B-A10B
    76.1%
    i
  7. MAI-Code-1-Flash
    75.0%
    i
  8. Qwen3.6 Plus
    74.2%
    i
  9. Nemotron 3 Super (120B A12B)
    72.6%
    i
  10. Mercury 2
    71.0%
    i
  11. Nova 2 Lite
    70.8%
    i
  12. Qwen3.5-35B-A3B
    70.2%
    i
  13. MiniMax M2.1
    70.0%
    i
  14. GPT OSS 120B High
    69.5%
    i
  15. MAI-Thinking-1
    69.0%
    i
  16. Mistral Medium 3.5
    69.0%
    i
  17. Nova 2 Omni
    68.7%
    i
  18. K-EXAONE-236B-A23B
    67.3%
    i
  19. Qwen3.5-9B
    64.5%
    i
  20. Qwen3.5-4B
    59.2%
    i