Benchmarks / Berkeley Function Calling Leaderboard (BFCL) V4

Reported by Berkeley Function Calling Leaderboard (BFCL) V4

Berkeley Function Calling Leaderboard (BFCL) V4

The same task on functions and questions contributed by real users.

Data older than 90 days: last updated 16 Dec 2025

Results dated
16 Dec 2025
Results
109 configurations of 81 models
Unit
% correct
Licence
Apache License 2.0

Berkeley Function Calling Leaderboard (BFCL) V4 in brief

What it measures. Berkeley Function Calling Leaderboard (BFCL) V4 is a third-party benchmark, shown here with credit. Measured here: The same task on functions and questions contributed by real users.

Latest results. , 81 models ranked.

Top 3 on Single-turn calls (user-contributed) (% correct, higher is better, each model at its best setting):

  1. BitAgent Bounty 8B (Bittensor): 93.1%
  2. Gemini 3 Pro Preview (Google): 83.1%
  3. Qwen3-32B (Alibaba): 82.0%

How to cite. Berkeley Function Calling Leaderboard (BFCL) V4, via Spring Prompt, results dated 16 December 2025. https://springprompt.com/benchmarks/bfcl These results belong to Berkeley Function Calling Leaderboard (BFCL) V4, not to us. Licence: Apache License 2.0. Credit Berkeley Function Calling Leaderboard (BFCL) V4 first.

Single-turn calls (user-contributed): Palmyra X 004

Top 15 of 81 results · % correct, higher is better. Choose a model to highlight it.Clear highlight

  1. 1 BitAgent Bounty 8BMode not statedBittensor 93.1%
  2. 2 Gemini 3 Pro PreviewPromptingGoogle 83.1%
  3. 3 Qwen3-32BFunction callingAlibaba 82.0%
  4. 4 Mistral Large 2.1 (2411)Function callingMistral AI 81.9%
  5. 5 Claude Sonnet 4.5Function callingAnthropic 81.1%
  6. 6 GLM-4.6 (reasoning on)Function callingZ.ai 80.9%
  7. 7 Nova 2 LiteFunction callingAmazon 80.8%
  8. 8 Arch-Agent-32BMode not statedKatanemo 80.7%
  9. 9 Qwen3-8BFunction callingAlibaba 80.5%
  10. 10 Qwen3-14BFunction callingAlibaba 80.0%
  11. 11 Claude Opus 4.5Function callingAnthropic 79.8%
  12. 12 Nanbeige4-3B-Thinking (2511)Function callingNanbeige 79.4%
  13. 13 Mistral Small 3.2 24BPromptingMistral AI 79.1%
  14. 14 GPT-4.1PromptingOpenAI 78.9%
  15. 15 Qwen3-235B-A22B-Instruct-2507PromptingAlibaba 78.7%
  16. 24 Palmyra X 004Function callingWriter 77.9%

Full results

BFCL: single-turn calls (user-contributed), % correct, higher is better
#ModelSingle-turn calls (user-contributed)
% correct, higher is better
Price
$ per million tokens, in / out
1 BitAgent Bounty 8BMode not statedBittensor
93.1%
–
2 Gemini 3 Pro PreviewPromptingGoogle · best of 2 settings
83.1%
–
3 Qwen3-32BFunction callingAlibaba · best of 2 settings
82.0%
$0.14 / $0.40
4 Mistral Large 2.1 (2411)Function callingMistral AI · best of 2 settings
81.9%
–
5 Claude Sonnet 4.5Function callingAnthropic · best of 2 settings
81.1%
$3 / $15
6 GLM-4.6 (reasoning on)Function callingZ.ai
80.9%
$0.50 / $2
7 Nova 2 LiteFunction callingAmazon
80.8%
$0.30 / $2.50
8 Arch-Agent-32BMode not statedKatanemo
80.7%
–
9 Qwen3-8BFunction callingAlibaba · best of 2 settings
80.5%
–
10 Qwen3-14BFunction callingAlibaba · best of 2 settings
80.0%
$0.10 / $0.22
11 Claude Opus 4.5Function callingAnthropic · best of 2 settings
79.8%
$5 / $25
12 Nanbeige4-3B-Thinking (2511)Function callingNanbeige
79.4%
–
13 Mistral Small 3.2 24BPromptingMistral AI · best of 2 settings
79.1%
$0.094 / $0.25
14 GPT-4.1PromptingOpenAI · best of 2 settings
78.9%
$2 / $8
15 Qwen3-235B-A22B-Instruct-2507PromptingAlibaba · best of 2 settings
78.7%
$0.15 / $0.75
15 Claude Haiku 4.5Function callingAnthropic · best of 2 settings
78.7%
$1 / $5
15 Kimi K2 (0711)Function callingMoonshot AI
78.7%
$0.57 / $2.30
18 Command A ReasoningFunction callingCohere
78.6%
–
19 Nova Pro 1.0Function callingAmazon
78.5%
$0.80 / $3.20
19 Command AFunction callingCohere
78.5%
$2.50 / $10
19 Grok 4.1 Fast (reasoning)Function callingxAI
78.5%
–
22 Qwen3-30B-A3B-Instruct-2507PromptingAlibaba · best of 2 settings
78.4%
$0.09 / $0.30
23 Gemini 2.5 FlashPromptingGoogle · best of 2 settings
78.2%
$0.30 / $2.50
24 Grok 4.1 Fast (non-reasoning)Function callingxAI
77.9%
–
24 Palmyra X 004Function callingWriter
77.9%
–
26 ToolACE-2-8BFunction callingHuawei Noah's Ark Lab and USTC
77.4%
–
27 Llama 3.3 70B InstructFunction callingMeta
76.6%
$0.71 / $0.71
28 Qwen3-4B-Instruct-2507Function callingAlibaba · best of 2 settings
76.4%
–
29 DeepSeek-V3.2-Exp (reasoning on)PromptingDeepSeek · best of 2 settings
76.0%
$0.27 / $0.41
30 Grok 4 (0709)Function callingxAI · best of 2 settings
75.6%
–
31 xLAM-2-32b-fc-rFunction callingSalesforce
75.5%
–
32 Falcon3 10B InstructFunction callingTII
75.4%
–
33 GPT-4.1 miniPromptingOpenAI · best of 2 settings
74.8%
$0.40 / $1.60
34 Llama 4 ScoutFunction callingMeta
74.7%
$0.10 / $0.30
35 Qwen3-1.7BFunction callingAlibaba
74.6%
–
36 Gemma 3 27BPromptingGoogle
74.5%
$0.08 / $0.35
37 Gemma 3 12BPromptingGoogle
74.2%
$0.05 / $0.15
38 Mistral NemoPromptingMistral AI · best of 2 settings
74.0%
$0.029 / $0.03
39 Llama 4 Maverick 17B 128E Instruct (FP8)Function callingMeta
73.7%
–
40 o3PromptingOpenAI · best of 2 settings
73.2%
$2 / $8
41 Arch-Agent-3BMode not statedKatanemo
72.9%
–
42 xLAM-2-70b-fc-rFunction callingSalesforce
72.2%
–
43 Llama 3.1 8B InstructPromptingMeta
70.8%
$0.05 / $0.08
43 o4-miniPromptingOpenAI · best of 2 settings
70.8%
–
45 GPT-5 nanoPromptingOpenAI · best of 2 settings
70.7%
$0.05 / $0.40
46 Hammer2.1-3BFunction callingMadeAgents
70.5%
–
47 GPT-5.2Function callingOpenAI · best of 2 settings
70.4%
$1.75 / $14
48 Nanbeige3.5-Pro-ThinkingFunction callingNanbeige
70.0%
–
49 Hammer2.1-1.5BFunction callingMadeAgents
69.5%
–
49 Hammer2.1-7BFunction callingMadeAgents
69.5%
–
51 Command R7B (12-2024)Function callingCohere
69.1%
$0.037 / $0.15
52 Falcon3 7B InstructFunction callingTII
68.3%
–
53 Mistral Medium 3Function callingMistral AI · best of 2 settings
68.0%
$0.40 / $2
53 xLAM-2-8b-fc-rFunction callingSalesforce
68.0%
–
55 Bielik 11B v2.3 InstructPromptingSpeakLeash and ACK Cyfronet AGH
67.8%
–
56 Arch-Agent-1.5BMode not statedKatanemo
67.7%
–
57 CoALM 70BMode not statedUIUC and Oumi
67.3%
–
58 CoALM 8BMode not statedUIUC and Oumi
66.8%
–
59 Nova Micro 1.0Function callingAmazon
66.3%
$0.035 / $0.14
60 Gemini 2.5 Flash-LiteFunction callingGoogle · best of 2 settings
65.8%
$0.10 / $0.40
61 MiniCPM3-4BFunction callingOpenBMB · best of 2 settings
65.2%
–
62 xLAM-2-3b-fc-rFunction callingSalesforce
62.9%
–
63 GPT-5 miniPromptingOpenAI · best of 2 settings
62.6%
$0.25 / $2
64 Gemma 3 4BPromptingGoogle
60.8%
$0.04 / $0.05
64 GPT-4.1 nanoFunction callingOpenAI · best of 2 settings
60.8%
$0.10 / $0.40
66 Phi-4PromptingMicrosoft
60.7%
$0.07 / $0.14
67 Granite 3.1 8B InstructFunction callingIBM
60.3%
–
67 Granite 3.2 8B InstructFunction callingIBM
60.3%
–
69 Granite 20B FunctionCallingFunction callingIBM
58.7%
–
70 Llama 3.2 3B InstructFunction callingMeta
58.3%
$0.05 / $0.33
71 Qwen3-0.6BFunction callingAlibaba · best of 2 settings
56.6%
–
72 xLAM-2-1b-fc-rFunction callingSalesforce
55.1%
–
73 Hammer2.1-0.5BFunction callingMadeAgents
54.6%
–
74 Falcon3 3B InstructFunction callingTII
54.5%
–
75 RZN-TPromptingPhronetic AI
49.7%
–
76 Granite 4.0 350MFunction callingIBM
46.1%
–
77 Gemma 3 1BPromptingGoogle
11.8%
–
77 Llama 3.2 1B InstructFunction callingMeta
11.8%
$0.027 / $0.20
79 Falcon3 1B InstructFunction callingTII
2.9%
–
80 Ministral 8B (2410)Function callingMistral AI
0.0%
–
80 Llama 3.1 Nemotron Ultra 253B v1Function callingNVIDIA
0.0%
–

Swipe the table sideways for more columns.

Source: Berkeley Function Calling Leaderboard (BFCL) V4, via Spring Prompt, springprompt.com/benchmarks/bfcl. Test your own prompt on these models →

Ranks follow the score as shown, so equal numbers share a rank. Each model is shown at its best setting; show every setting. Results as published by Berkeley Function Calling Leaderboard (BFCL) V4; we do not re-run them.

Progress over time

Single-turn calls (user-contributed), plotted at each model's release date, higher is better. Each model counts once, at its best setting. Scores are today's, not launch day's; 38 models without a known release date are left off. Hover or tap a mark for its model.

Best score so far

Every model at its release date. The line moves each time a new release beats every earlier one, and each new best is named where there is room.

0%50%100% Jul 20242025Jul2026Jul Mistral NemoMistral Large 2.1 (2411)Qwen3-32BGemini 3 Pro Preview

Best score by lab

Each lab's best so far, stepping whenever one of its releases beats its own earlier models. The 6 labs with the best scores today; 7 others not drawn.

GoogleAlibabaMistral AIAnthropicZ.aiAmazon

70%75%80%85% Jul 20242025Jul2026Jul GoogleAlibabaMistral AIAnthropicZ.aiAmazon

What it measures

The same task on functions and questions contributed by real users.

What it does not measure

Not reliability on your own tools: BFCL's functions and queries are a fixed test set, and a correct call is judged by its form, not by what it achieved.

Berkeley Function Calling Leaderboard (BFCL) V4 by the UC Berkeley Gorilla team, Apache License 2.0.