Skip to content

Berkeley Function Calling Leaderboard v4

A function-calling benchmark for tool selection, schema adherence, and argument correctness.

19 models

Higher is better.

  1. 1Atria Dawn PreviewShanghai Artificial Intelligence Laboratory77.0
  2. 2Qwen3.7 MaxAlibaba75.0
  3. 3BTL-4Bad Theory Labs73.5
  4. 4Ling 3.0 FlashInclusionAI73.0
  5. 5Qwen3.7 PlusAlibaba72.9
  6. 6Pokee-Isaac 28BPokee AI70.9
  7. 7Kimi K2.5Moonshot AI68.3
  8. 8MiniCPM5-2BOpenBMB66.6
  9. 9LFM2.5-24B-A2BLiquid AI65.0
  10. 10LLaDA2.2-flashInclusionAI60.8
  11. 11LFM2.5-2.6BLiquid AI56.9
  12. 12Granite 4.2 8BIBM52.4
  13. 13LFM2.5-8B-A1BLiquid AI49.7
  14. 14Mellum2-12B-A2.5B-InstructJetBrains44.2
  15. 15ZAYA1-8BZyphra39.2
  16. 16LFM2.5-VL-3BLiquid AI32.5
  17. 17MiniCPM5-1BOpenBMB25.1
  18. 18LFM2.5-VL-450MLiquid AI21.1
  19. 19LFM2.5-230MLiquid AI21.0
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Tool invocation and schema evaluation
Tasks
Function-calling tasks
Authors
Shishir G. Patil et al. (2025)

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