Wikibench

BFCL-V4

agents

Berkeley Function Calling Leaderboard V4 (BFCL-V4) evaluates LLMs on their ability to accurately call functions and APIs, including simple, multiple, parallel, and nested function calls across diverse programming scenarios.

Leaderboard

Showing 12 of 12 results

  1. Qwen3.7 Max
    75.0%
    i
  2. Qwen3.5-397B-A17B
    72.9%
    i
  3. Qwen3.5-122B-A10B
    72.2%
    i
  4. Qwen3.5-27B
    68.5%
    i
  5. Qwen3.5-35B-A3B
    67.3%
    i
  6. Qwen3.5-9B
    66.1%
    i
  7. Nova 2 Pro
    61.6%
    i
  8. Nova 2 Lite
    60.3%
    i
  9. Nova 2 Omni
    58.3%
    i
  10. Qwen3.5-4B
    50.3%
    i
  11. Qwen3.5-2B
    43.6%
    i
  12. Qwen3.5-0.8B
    25.3%
    i