Skip to content

Benchmark

BFCL Relevance

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
0.25

Score distribution

77.3%
Mean
81.3%
Median
18.5%
Std dev
50%
10th percentile
93.8%
90th percentile
0%
Min
0%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.25 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
1Gemma 3 4B ITPrompt· Berkeley Function-Calling Leaderboard data100%
2GPT 4.1Prompt· Berkeley Function-Calling Leaderboard data100%
3Llama 3.3 70B Instructfc· Berkeley Function-Calling Leaderboard data100%
4Llama 4 Maverickfc· Berkeley Function-Calling Leaderboard data100%
5Llama 4 Scoutfc· Berkeley Function-Calling Leaderboard data100%
6DeepSeek V3.2 ExpPrompt + Thinking· Berkeley Function-Calling Leaderboard data93.8%
7Gemma 3 12B ITPrompt· Berkeley Function-Calling Leaderboard data93.8%
8GPT-4.1 nanofc· Berkeley Function-Calling Leaderboard data93.8%
9GPT-5 miniPrompt· Berkeley Function-Calling Leaderboard data93.8%
10GPT-5 nanoPrompt· Berkeley Function-Calling Leaderboard data93.8%
11Llama 3.1 8B InstructPrompt· Berkeley Function-Calling Leaderboard data93.8%
12Mistral Largefc· Berkeley Function-Calling Leaderboard data93.8%
13Mistral LargePrompt· Berkeley Function-Calling Leaderboard data93.8%
14Mistral SmallPrompt· Berkeley Function-Calling Leaderboard data93.8%
15Nova Profc· Berkeley Function-Calling Leaderboard data93.8%
16Open Mistral NemoPrompt· Berkeley Function-Calling Leaderboard data93.8%
17OpenAI o3Prompt· Berkeley Function-Calling Leaderboard data93.8%
18Qwen3 235B A22B Instruct 2507Prompt· Berkeley Function-Calling Leaderboard data93.8%
19Qwen3 30B A3B InstructPrompt· Berkeley Function-Calling Leaderboard data93.8%
20Qwen3 32Bfc· Berkeley Function-Calling Leaderboard data93.8%
21Qwen3 8Bfc· Berkeley Function-Calling Leaderboard data93.8%
22GPT 4.1fc· Berkeley Function-Calling Leaderboard data87.5%
23GPT-4.1 miniPrompt· Berkeley Function-Calling Leaderboard data87.5%
24Grok 4fc· Berkeley Function-Calling Leaderboard data87.5%
25Llama 3.2 3Bfc· Berkeley Function-Calling Leaderboard data87.5%
26Mistral Smallfc· Berkeley Function-Calling Leaderboard data87.5%
27Qwen3 14Bfc· Berkeley Function-Calling Leaderboard data87.5%
28Qwen3 235B A22B Instruct 2507fc· Berkeley Function-Calling Leaderboard data87.5%
29Qwen3 4B Instruct 2507fc· Berkeley Function-Calling Leaderboard data87.5%
30Qwen3 4B Instruct 2507Prompt· Berkeley Function-Calling Leaderboard data87.5%
31Gemma 3 27B ITPrompt· Berkeley Function-Calling Leaderboard data81.3%
32GPT-4.1 minifc· Berkeley Function-Calling Leaderboard data81.3%
33Grok 4Prompt· Berkeley Function-Calling Leaderboard data81.3%
34Grok 4.1 Fastfc· Berkeley Function-Calling Leaderboard data81.3%
35Nova Microfc· Berkeley Function-Calling Leaderboard data81.3%
36o4 Minifc· Berkeley Function-Calling Leaderboard data81.3%
37o4 MiniPrompt· Berkeley Function-Calling Leaderboard data81.3%
38Open Mistral Nemofc· Berkeley Function-Calling Leaderboard data81.3%
39OpenAI o3fc· Berkeley Function-Calling Leaderboard data81.3%
40Qwen3 1.7Bfc· Berkeley Function-Calling Leaderboard data81.3%
41Qwen3 14BPrompt· Berkeley Function-Calling Leaderboard data81.3%
42Qwen3 30B A3B Instructfc· Berkeley Function-Calling Leaderboard data81.3%
43Qwen3 32BPrompt· Berkeley Function-Calling Leaderboard data81.3%
44Gemini 2.5 Flashfc· Berkeley Function-Calling Leaderboard data75%
45Gemini 3 Pro Previewfc· Berkeley Function-Calling Leaderboard data75%
46GLM 4.6FC thinking· Berkeley Function-Calling Leaderboard data75%
47GPT 5.2fc· Berkeley Function-Calling Leaderboard data75%
48GPT 5.2Prompt· Berkeley Function-Calling Leaderboard data75%
49GPT-5 nanofc· Berkeley Function-Calling Leaderboard data75%
50Kimi K2 Instructfc· Berkeley Function-Calling Leaderboard data75%

Showing 1–50 of 76 results.

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