Benchmarks / Berkeley Function Calling Leaderboard (BFCL) V4

Reported by Berkeley Function Calling Leaderboard (BFCL) V4

Berkeley Function Calling Leaderboard (BFCL) V4

Making a call when a suitable function is available.

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: Making a call when a suitable function is available.

Latest results. , 109 configurations ranked.

Top 3 on Relevance detection (% correct, higher is better):

  1. Gemma 3 4B (Google): 100.0%
  2. Llama 3.3 70B Instruct (Meta): 100.0%
  3. Llama 4 Maverick 17B 128E Instruct (FP8) (Meta): 100.0%

The first 7 are too close to tell apart, so read the order among them as a tie.

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.

Relevance detection: Qwen3-4B-Instruct-2507

Top 16 of 109 results · % correct, higher is better · ≈ cannot be told apart from the leader. Choose a model to highlight it.Clear highlight

  1. 1≈ Gemma 3 4BPromptingGoogle 100.0%
  2. 1≈ Llama 3.3 70B InstructFunction callingMeta 100.0%
  3. 1≈ Llama 4 Maverick 17B 128E Instruct (FP8)Function callingMeta 100.0%
  4. 1≈ Llama 4 ScoutFunction callingMeta 100.0%
  5. 1≈ Nanbeige3.5-Pro-ThinkingFunction callingNanbeige 100.0%
  6. 1≈ GPT-4.1PromptingOpenAI 100.0%
  7. 1≈ Falcon3 7B InstructFunction callingTII 100.0%
  8. 8 Qwen3-235B-A22B-Instruct-2507PromptingAlibaba 93.8%
  9. 8 Qwen3-30B-A3B-Instruct-2507PromptingAlibaba 93.8%
  10. 8 Qwen3-32BFunction callingAlibaba 93.8%
  11. 8 Qwen3-8BFunction callingAlibaba 93.8%
  12. 8 Nova Pro 1.0Function callingAmazon 93.8%
  13. 8 DeepSeek-V3.2-Exp (reasoning on)PromptingDeepSeek 93.8%
  14. 8 Gemma 3 12BPromptingGoogle 93.8%
  15. 8 Llama 3.1 8B InstructPromptingMeta 93.8%
  16. 27 Qwen3-4B-Instruct-2507Function callingAlibaba 87.5%
  17. 27 Qwen3-4B-Instruct-2507PromptingAlibaba 87.5%

Full results

BFCL: relevance detection, % correct, higher is better
#ModelRelevance detection
% correct, higher is better
Price
$ per million tokens, in / out
1≈ Gemma 3 4BPromptingGoogle
100.0%
$0.04 / $0.05
1≈ Llama 3.3 70B InstructFunction callingMeta
100.0%
$0.71 / $0.71
1≈ Llama 4 Maverick 17B 128E Instruct (FP8)Function callingMeta
100.0%
–
1≈ Llama 4 ScoutFunction callingMeta
100.0%
$0.10 / $0.30
1≈ Nanbeige3.5-Pro-ThinkingFunction callingNanbeige
100.0%
–
1≈ GPT-4.1PromptingOpenAI
100.0%
$2 / $8
1≈ Falcon3 7B InstructFunction callingTII
100.0%
–
8 Qwen3-235B-A22B-Instruct-2507PromptingAlibaba
93.8%
$0.15 / $0.75
8 Qwen3-30B-A3B-Instruct-2507PromptingAlibaba
93.8%
$0.09 / $0.30
8 Qwen3-32BFunction callingAlibaba
93.8%
$0.14 / $0.40
8 Qwen3-8BFunction callingAlibaba
93.8%
–
8 Nova Pro 1.0Function callingAmazon
93.8%
$0.80 / $3.20
8 DeepSeek-V3.2-Exp (reasoning on)PromptingDeepSeek
93.8%
$0.27 / $0.41
8 Gemma 3 12BPromptingGoogle
93.8%
$0.05 / $0.15
8 Llama 3.1 8B InstructPromptingMeta
93.8%
$0.05 / $0.08
8 Mistral Large 2.1 (2411)Function callingMistral AI
93.8%
–
8 Mistral Large 2.1 (2411)PromptingMistral AI
93.8%
–
8 Mistral NemoPromptingMistral AI
93.8%
$0.029 / $0.03
8 Mistral Small 3.2 24BPromptingMistral AI
93.8%
$0.094 / $0.25
8 GPT-4.1 nanoFunction callingOpenAI
93.8%
$0.10 / $0.40
8 GPT-5 miniPromptingOpenAI
93.8%
$0.25 / $2
8 GPT-5 nanoPromptingOpenAI
93.8%
$0.05 / $0.40
8 o3PromptingOpenAI
93.8%
$2 / $8
8 Bielik 11B v2.3 InstructPromptingSpeakLeash and ACK Cyfronet AGH
93.8%
–
8 Falcon3 10B InstructFunction callingTII
93.8%
–
8 CoALM 70BMode not statedUIUC and Oumi
93.8%
–
27 Qwen3-14BFunction callingAlibaba
87.5%
$0.10 / $0.22
27 Qwen3-235B-A22B-Instruct-2507Function callingAlibaba
87.5%
$0.15 / $0.75
27 Qwen3-4B-Instruct-2507Function callingAlibaba
87.5%
–
27 Qwen3-4B-Instruct-2507PromptingAlibaba
87.5%
–
27 Granite 20B FunctionCallingFunction callingIBM
87.5%
–
27 Llama 3.2 3B InstructFunction callingMeta
87.5%
$0.05 / $0.33
27 Mistral Small 3.2 24BFunction callingMistral AI
87.5%
$0.094 / $0.25
27 GPT-4.1 miniPromptingOpenAI
87.5%
$0.40 / $1.60
27 GPT-4.1Function callingOpenAI
87.5%
$2 / $8
27 xLAM-2-1b-fc-rFunction callingSalesforce
87.5%
–
27 xLAM-2-3b-fc-rFunction callingSalesforce
87.5%
–
27 xLAM-2-8b-fc-rFunction callingSalesforce
87.5%
–
27 CoALM 8BMode not statedUIUC and Oumi
87.5%
–
27 Grok 4 (0709)Function callingxAI
87.5%
–
41 Qwen3-1.7BFunction callingAlibaba
81.3%
–
41 Qwen3-14BPromptingAlibaba
81.3%
$0.10 / $0.22
41 Qwen3-30B-A3B-Instruct-2507Function callingAlibaba
81.3%
$0.09 / $0.30
41 Qwen3-32BPromptingAlibaba
81.3%
$0.14 / $0.40
41 Nova Micro 1.0Function callingAmazon
81.3%
$0.035 / $0.14
41 Command AFunction callingCohere
81.3%
$2.50 / $10
41 Gemma 3 27BPromptingGoogle
81.3%
$0.08 / $0.35
41 Granite 4.0 350MFunction callingIBM
81.3%
–
41 Arch-Agent-32BMode not statedKatanemo
81.3%
–
41 Mistral NemoFunction callingMistral AI
81.3%
$0.029 / $0.03
41 GPT-4.1 miniFunction callingOpenAI
81.3%
$0.40 / $1.60
41 o3Function callingOpenAI
81.3%
$2 / $8
41 o4-miniFunction callingOpenAI
81.3%
–
41 o4-miniPromptingOpenAI
81.3%
–
41 RZN-TPromptingPhronetic AI
81.3%
–
41 xLAM-2-32b-fc-rFunction callingSalesforce
81.3%
–
41 Falcon3 3B InstructFunction callingTII
81.3%
–
41 Palmyra X 004Function callingWriter
81.3%
–
41 Grok 4 (0709)PromptingxAI
81.3%
–
41 Grok 4.1 Fast (non-reasoning)Function callingxAI
81.3%
–
41 Grok 4.1 Fast (reasoning)Function callingxAI
81.3%
–
62 Qwen3-0.6BFunction callingAlibaba
75.0%
–
62 Qwen3-0.6BPromptingAlibaba
75.0%
–
62 Qwen3-8BPromptingAlibaba
75.0%
–
62 Nova 2 LiteFunction callingAmazon
75.0%
$0.30 / $2.50
62 Gemini 2.5 FlashFunction callingGoogle
75.0%
$0.30 / $2.50
62 Gemini 3 Pro PreviewFunction callingGoogle
75.0%
–
62 ToolACE-2-8BFunction callingHuawei Noah's Ark Lab and USTC
75.0%
–
62 Granite 3.2 8B InstructFunction callingIBM
75.0%
–
62 Arch-Agent-1.5BMode not statedKatanemo
75.0%
–
62 Hammer2.1-1.5BFunction callingMadeAgents
75.0%
–
62 Mistral Medium 3Mode not statedMistral AI
75.0%
$0.40 / $2
62 Kimi K2 (0711)Function callingMoonshot AI
75.0%
$0.57 / $2.30
62 Nanbeige4-3B-Thinking (2511)Function callingNanbeige
75.0%
–
62 GPT-5.2Function callingOpenAI
75.0%
$1.75 / $14
62 GPT-5.2PromptingOpenAI
75.0%
$1.75 / $14
62 GPT-5 nanoFunction callingOpenAI
75.0%
$0.05 / $0.40
62 xLAM-2-70b-fc-rFunction callingSalesforce
75.0%
–
62 GLM-4.6 (reasoning on)Function callingZ.ai
75.0%
$0.50 / $2
80 Claude Opus 4.5PromptingAnthropic
68.8%
$5 / $25
80 Claude Sonnet 4.5Function callingAnthropic
68.8%
$3 / $15
80 BitAgent Bounty 8BMode not statedBittensor
68.8%
–
80 Command A ReasoningFunction callingCohere
68.8%
–
80 Command R7B (12-2024)Function callingCohere
68.8%
$0.037 / $0.15
80 Gemini 3 Pro PreviewPromptingGoogle
68.8%
–
80 Granite 3.1 8B InstructFunction callingIBM
68.8%
–
80 Arch-Agent-3BMode not statedKatanemo
68.8%
–
80 Hammer2.1-0.5BFunction callingMadeAgents
68.8%
–
80 GPT-4.1 nanoPromptingOpenAI
68.8%
$0.10 / $0.40
80 MiniCPM3-4BFunction callingOpenBMB
68.8%
–
91 Claude Haiku 4.5Function callingAnthropic
62.5%
$1 / $5
91 Claude Opus 4.5Function callingAnthropic
62.5%
$5 / $25
91 Gemini 2.5 FlashPromptingGoogle
62.5%
$0.30 / $2.50
91 Mistral Medium 3Function callingMistral AI
62.5%
$0.40 / $2
91 GPT-5 miniFunction callingOpenAI
62.5%
$0.25 / $2
96 Hammer2.1-3BFunction callingMadeAgents
56.3%
–
96 MiniCPM3-4BPromptingOpenBMB
56.3%
–
98 Gemini 2.5 Flash-LitePromptingGoogle
50.0%
$0.10 / $0.40
98 Hammer2.1-7BFunction callingMadeAgents
50.0%
–
98 Phi-4PromptingMicrosoft
50.0%
$0.07 / $0.14
101 Gemini 2.5 Flash-LiteFunction callingGoogle
43.8%
$0.10 / $0.40
101 Llama 3.2 1B InstructFunction callingMeta
43.8%
$0.027 / $0.20
103 Claude Sonnet 4.5PromptingAnthropic
37.5%
$3 / $15
103 DeepSeek-V3.2-ExpFunction callingDeepSeek
37.5%
$0.27 / $0.41
103 Gemma 3 1BPromptingGoogle
37.5%
–
106 Claude Haiku 4.5PromptingAnthropic
31.3%
$1 / $5
107 Ministral 8B (2410)Function callingMistral AI
0.0%
–
107 Llama 3.1 Nemotron Ultra 253B v1Function callingNVIDIA
0.0%
–
107 Falcon3 1B InstructFunction callingTII
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. Results as published by Berkeley Function Calling Leaderboard (BFCL) V4; we do not re-run them.

Progress over time

Relevance detection, 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.

-50%0%50%100%150% Jul 20242025Jul2026Jul Mistral NemoLlama 3.3 70B Instruct

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.

MetaGoogleOpenAIMistral AIAmazonAlibaba

92%94%96%98%100%102% Jul 20242025Jul2026Jul MetaGoogleOpenAIMistral AIAmazonAlibaba

What it measures

Making a call when a suitable function is available.

What it does not measure

Not whether the call itself was right; only that one was attempted.

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