BFCL Overall

Berkeley Function Calling Leaderboard overall accuracy.

Published by

Berkeley Function Calling Leaderboard

Category

Agents & tools

Index weight

1.0

Models

83

Data as of

9 Sept 2026

Apache 2.0 — ShishirPatil/gorilla (BFCL).

1. 1Claude Opus 4.577.5%
2. 2Claude Sonnet 4.573.2%
3. 3Gemini 3 Pro72.5%
4. 4GLM-4.672.4%
5. 5Grok 4.169.6%
6. 6Claude Haiku 4.568.7%
7. 7o363.0%
8. 8Grok 463.0%
9. 9Kimi K259.1%
10. 10Grok 4.158.3%
11. 11Command A57.1%
12. 12DeepSeek V3.256.7%
13. 13Gemini 2.5 Flash56.2%
14. 14GPT-5.255.9%
15. 15GPT-5 mini55.5%

SearchAll providersAlibabaAmazonAnthropicBittensorCohereDeepSeekGoogleHuawei Noah & USTCIbmIBMkatanemoMadeAgentsMetaMicrosoftMistral AIMoonshot AINanbeigeNVIDIAOpenAIopenbmbPhronetic AIQwenSalesforceSpeakLeash & ACK Cyfronet AGHTII UAEUIUC + OumiWriterxAIZhipu AIOpen weights83 of 83

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	Claude Opus 4.5Anthropic	77.5%	–	58.0	–
2	Claude Sonnet 4.5Anthropic	73.2%	–	52.3	–
3	Gemini 3 ProGoogle	72.5%	–	61.6	–
4	GLM-4.6Zhipu AIopen	72.4%	–	52.3	–
5	Grok 4.1xAI	69.6%	–	55.9	–
6	Claude Haiku 4.5Anthropic	68.7%	–	47.9	–
7	o3OpenAI	63.0%	–	60.3	–
8	Grok 4xAI	63.0%	prompt, 0709	57.2	–
9	Kimi K2Moonshot AIopen	59.1%	–	48.8	–
10	Grok 4.1xAI	58.3%	–	47.9	–
11	Command ACohereopen	57.1%	–	–	–
12	DeepSeek V3.2DeepSeekopen	56.7%	–	55.9	–
13	Gemini 2.5 FlashGoogle	56.2%	–	48.8	–
14	GPT-5.2OpenAI	55.9%	–	61.5	–
15	GPT-5 miniOpenAI	55.5%	–	55.8	–
16	SXlam 2 32B Fc RSalesforce	54.7%	–	–	–
17	DeepSeek V3.2DeepSeekopen	54.1%	–	54.7	–
18	GPT-4.1OpenAI	54.0%	–	47.3	–
19	o4-miniOpenAI	53.2%	–	54.9	–
20	SXlam 2 70B Fc RSalesforce	53.1%	–	–	–
21	Qwen3 235B A22BAlibabaopen	52.1%	prompt, 2507	48.7	–
22	GPT-5 nanoOpenAI	51.5%	–	51.1	–
23	NNanbeige4 3BNanbeige	51.4%	thinking, fc, 2511	–	–
24	GPT-4.1 miniOpenAI	50.5%	–	44.5	–
25	Qwen3 32BAlibabaopen	48.7%	–	45.9	–
26	NNanbeige3 5 ProNanbeige	47.7%	–	–	–
27	SXlam 2 8B Fc RSalesforce	46.7%	–	–	–
28	Command ACohereopen	46.5%	–	42.1	–
29	BBitagent Bounty 8BBittensor	46.2%	–	–	–
30	KArch Agent 32Bkatanemo	45.4%	–	–	–
31	Qwen3 8BAlibabaopen	42.6%	–	38.9	–
32	HToolace 2 8BHuawei Noah & USTC	42.4%	–	–	–
33	Qwen3 30B A3BAlibabaopen	41.4%	fc, 2507	45.4	–
34	SXlam 2 3B Fc RSalesforce	41.2%	–	–	–
35	Qwen3 14BAlibabaopen	41.0%	–	40.7	–
36	Mistral Large 2Mistral AIopen	38.4%	fc, 2411	41.2	–
37	Mistral Medium (latest)Mistral AI	37.7%	medium, 2505	45.8	–
38	Llama 4 MaverickMetaopen	37.3%	–	42.6	–
39	Mistral Small (latest)Mistral AI	37.1%	fc, 2506	41.1	–
40	Gemini 2.5 Flash-LiteGoogle	36.9%	–	41.6	–
41	qwen3-4b-instruct-2507Alibabaopen	35.7%	fc, 2507	41.7	–
42	KArch Agent 3Bkatanemo	35.4%	–	–	–
43	GPT-4.1 nanoOpenAI	33.0%	–	38.1	–
44	KArch Agent 1 5Bkatanemo	32.1%	–	–	–
45	Command R7bCohere	32.1%	–	–	–
46	Llama 3.3 70BMetaopen	31.9%	–	40.7	–
47	MHammer2 1 7BMadeAgents	31.7%	–	–	–
48	SXlam 2 1B Fc RSalesforce	30.4%	–	–	–
49	Gemma 3 12B ITNVIDIAopen	30.4%	–	37.2	–
50	MHammer2 1 3BMadeAgents	29.7%	–	–	–
51	Gemma 3 27BGoogleopen	29.5%	–	39.7	–
52	Phi-4Microsoftopen	28.8%	–	39.7	–
53	Qwen3-1.7BAlibaba	28.4%	–	35.8	–
54	Llama 4 ScoutMetaopen	28.1%	–	39.7	–
55	UCoalm 70BUIUC + Oumi	28.0%	CoALM-70B	–	–
56	MHammer2 1.1 5BMadeAgents	27.9%	–	–	–
57	WPalmyra X 004Writer	27.9%	–	–	–
58	Open Mistral NemoMistral AIopen	27.6%	fc, 2407	–	–
59	Granite 3.1 8BIbm	27.1%	–	–	–
60	Nova 2 LiteAmazon	27.1%	–	–	–

Show 23 more