Inference

Berkeley Function Calling Leaderboard v4

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

  • In index
  • Agentic
ModelsModels with a result.22
Top resultBest result on this test.88.5%BTL-3
Top-3 spreadPoints from first to third.13.5 pts
YearYear of release.N/A

Result and price

020406080100$0.03$0.1$0.3$1$3$10RESULTOUTPUT PRICE PER 1M TOKENS · LOG SCALE

Results

22 results
#ModelResultThe score from the source.
01BTL-388.5%
02Atria Dawn Preview77.0%
03Qwen3.7 Max75.0%
04BTL-473.5%
05Ling 3.0 Flash73.0%
06Qwen3.7 Plus72.9%
07Pokee-Isaac 28B70.9%
08MiniCPM5-2B66.6%
09Granite 4.2 30B61.4%
10LLaDA2.2-flash60.8%
11LFM2.5-2.6B56.9%
12Granite 4.2 3B52.4%
13Granite 4.2 8B52.4%
14LFM2.5-8B-A1B49.7%
15LLaDA2.2-mini47.7%
16Mellum2-12B-A2.5B-Thinking45.6%
17Mellum2-12B-A2.5B-Instruct44.2%
18ZAYA1-8B39.2%
19LFM2.5-VL-3B32.5%
20MiniCPM5-1B25.1%
21LFM2.5-VL-450M21.1%
22LFM2.5-230M21.0%