BFCL v3 Leaderboard

Berkeley Function Calling Leaderboard v3 — testing function/tool calling accuracy.

Data from LayerLens

Models

Best Score

76.7

Average

57.8

Std Dev

17.7

Categories

Multi-turn

SourceLayerLens

Provider	Model	Input $/M	Output $/M	BFCL v3	Actions
Z Z AI	GLM 4.5 Thinking	$0.550	$2.000	76.7	Try
AL Alibaba	Qwen3 32B Thinking	$0.080	$0.240	75.7	Try
AL Alibaba	Qwen3 32B	$0.080	$0.240	75.7	Try
AL Alibaba	Qwen3 Max	$1.200	$6.000	74.9	Try
Z Z AI	GLM-4.7-Flash Thinking	$0.060	$0.400	74.6	Try
Z Z AI	GLM-4.7-Flash	$0.060	$0.400	74.6	Try
O OpenAI	GPT-4o	$2.500	$10.000	74.2	Try
Z Z AI	GLM 4.5 Air	$0.130	$0.850	69.1	Try
AM Amazon	Nova Pro 1.0	$0.800	$3.200	67.9	Try
K Kimi	Kimi K2.5 Thinking	$0.450	$2.200	64.5	Try
K Kimi	Kimi K2.5	$0.450	$2.200	64.5	Try
P Prime Intellect	INTELLECT-3	$0.200	$1.100	63.5	Try
M Meta	Llama 4 Scout	$0.080	$0.300	55.7	Try
G Google	Gemini 3 Flash Preview Thinking	$0.500	$3.000	53.5	Try
MM MiniMax	MiniMax M1	$0.400	$2.200	47.8	Try
NV NVIDIA	Nemotron 3 Nano 30B A3B Thinking	$0.050	$0.200	41.6	Try
NV NVIDIA	Nemotron 3 Nano 30B A3B	$0.050	$0.200	41.6	Try
MS Microsoft Azure	Phi 4	$0.060	$0.140	40.8	Try
A Anthropic	Claude Opus 4 Thinking	$15.000	$75.000	25.3	Try
A Anthropic	Claude Opus 4	$15.000	$75.000	25.3	Try
K Kimi	Kimi K2 0711	$0.550	$2.200	25.3	Try

Pricing from OpenRouter. Benchmarks from Artificial Analysis.

Deploy OpenClaw in Under 1 Minute— We handle hosting, scaling, and maintenance

93 out of our 301 tracked models have had a price change in March.

Get our weekly newsletter on pricing changes, new releases, and tools.

About BFCL v3

Berkeley Function Calling Leaderboard v3 — testing function/tool calling accuracy.

This leaderboard shows all models with BFCL v3 benchmark scores, ranked from highest to lowest. Pricing data is included to help you compare performance against cost.