Skip to content

Benchmark

BFCL Live

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages.

Models measured
49
Scores recorded
76
Sources
1
IQ weight
1.40

Score distribution

68.9%
Mean
74%
Median
15.2%
Std dev
54.9%
10th percentile
80.5%
90th percentile
0%
Min
0%83.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.40 for chat models, measured over 49 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Gemini 3 Pro PreviewPrompt· Berkeley Function-Calling Leaderboard data83.1%
2Qwen3 32Bfc· Berkeley Function-Calling Leaderboard data82%
3Qwen3 32BPrompt· Berkeley Function-Calling Leaderboard data82%
4Mistral Largefc· Berkeley Function-Calling Leaderboard data81.9%
5Gemini 3 Pro Previewfc· Berkeley Function-Calling Leaderboard data81.7%
6Claude Sonnet 4.5fc· Berkeley Function-Calling Leaderboard data81.1%
7GLM 4.6FC thinking· Berkeley Function-Calling Leaderboard data80.9%
8Nova 2 Litefc· Berkeley Function-Calling Leaderboard data80.8%
9Qwen3 8Bfc· Berkeley Function-Calling Leaderboard data80.5%
10Qwen3 8BPrompt· Berkeley Function-Calling Leaderboard data80.1%
11Qwen3 14Bfc· Berkeley Function-Calling Leaderboard data80%
12Claude Opus 4.5fc· Berkeley Function-Calling Leaderboard data79.8%
13Qwen3 14BPrompt· Berkeley Function-Calling Leaderboard data79.4%
14Mistral SmallPrompt· Berkeley Function-Calling Leaderboard data79.1%
15GPT 4.1Prompt· Berkeley Function-Calling Leaderboard data78.9%
16Claude Haiku 4.5fc· Berkeley Function-Calling Leaderboard data78.7%
17Kimi K2 Instructfc· Berkeley Function-Calling Leaderboard data78.7%
18Qwen3 235B A22B Instruct 2507Prompt· Berkeley Function-Calling Leaderboard data78.7%
19Command Afc· Berkeley Function-Calling Leaderboard data78.6%
20Nova Profc· Berkeley Function-Calling Leaderboard data78.5%
21Grok 4.1 Fastfc· Berkeley Function-Calling Leaderboard data78.5%
22Qwen3 30B A3B InstructPrompt· Berkeley Function-Calling Leaderboard data78.4%
23Gemini 2.5 FlashPrompt· Berkeley Function-Calling Leaderboard data78.2%
24Qwen3 30B A3B Instructfc· Berkeley Function-Calling Leaderboard data77.9%
25Mistral Smallfc· Berkeley Function-Calling Leaderboard data77.3%
26Llama 3.3 70B Instructfc· Berkeley Function-Calling Leaderboard data76.6%
27Qwen3 4B Instruct 2507fc· Berkeley Function-Calling Leaderboard data76.4%
28Claude Opus 4.5Prompt· Berkeley Function-Calling Leaderboard data76%
29DeepSeek V3.2 ExpPrompt + Thinking· Berkeley Function-Calling Leaderboard data76%
30Grok 4fc· Berkeley Function-Calling Leaderboard data75.6%
31GPT-4.1 miniPrompt· Berkeley Function-Calling Leaderboard data74.8%
32Llama 4 Scoutfc· Berkeley Function-Calling Leaderboard data74.7%
33Qwen3 4B Instruct 2507Prompt· Berkeley Function-Calling Leaderboard data74.7%
34Qwen3 1.7Bfc· Berkeley Function-Calling Leaderboard data74.6%
35Gemma 3 27B ITPrompt· Berkeley Function-Calling Leaderboard data74.5%
36Gemini 2.5 Flashfc· Berkeley Function-Calling Leaderboard data74.4%
37Gemma 3 12B ITPrompt· Berkeley Function-Calling Leaderboard data74.2%
38Open Mistral NemoPrompt· Berkeley Function-Calling Leaderboard data74%
39Open Mistral Nemofc· Berkeley Function-Calling Leaderboard data73.8%
40Llama 4 Maverickfc· Berkeley Function-Calling Leaderboard data73.7%
41OpenAI o3Prompt· Berkeley Function-Calling Leaderboard data73.2%
42Grok 4Prompt· Berkeley Function-Calling Leaderboard data72.5%
43Llama 3.1 8B InstructPrompt· Berkeley Function-Calling Leaderboard data70.8%
44o4 MiniPrompt· Berkeley Function-Calling Leaderboard data70.8%
45GPT-5 nanoPrompt· Berkeley Function-Calling Leaderboard data70.7%
46GPT 5.2fc· Berkeley Function-Calling Leaderboard data70.4%
47GPT 4.1fc· Berkeley Function-Calling Leaderboard data70%
48command-r7bfc· Berkeley Function-Calling Leaderboard data69.1%
49Qwen3 235B A22B Instruct 2507fc· Berkeley Function-Calling Leaderboard data68.9%
50GPT-4.1 minifc· Berkeley Function-Calling Leaderboard data68.8%

Showing 1–50 of 76 results.

Data credits: Berkeley Function-Calling Leaderboard (Gorilla, UC Berkeley), Apache-2.0 (licence)