Execution Accuracy on BFCL v2 benchmark leaderboard · SOTA2 Research

Execution Accuracy on BFCL v2

88.24Non-Live AST AccuracyQwen2.5-7B-Instruct-CAST

Evaluation Results

Method	Links
Qwen2.5-7B-Instruct-CAST2026.05		88.24	87.4	88.43
Qwen2.5-Coder-7B-Instruct-CAST2026.05		87.12	82.43	84.7
Qwen2.5-7B-Instruct-GRPO2026.05		87.05	80.29	83.67
GPT-4o2026.05		86.83	78.92	82.88
Qwen2.5-Coder-7B-Instruct-GRPO2026.05		86.51	75.07	80.79
DeepSeek-V32026.05		86.37	75.28	80.83
Qwen2.5-7B-Instruct-SFT2026.05		86.23	78.9	82.58
Granite2026.05		86.17	79.19	84.71
Qwen2.5-Coder-7B-Instruct-SFT2026.05		86.07	74.92	80.86
Gorilla2026.05		86.02	80.44	82.21
Qwen-Max2026.05		84.97	80.85	82.91
Llama-3.2-8B-Instruct-CAST2026.05		83.95	76.84	80.43
Llama-3.2-8B-Instruct-SFT2026.05		83.14	73.82	78.48
OTC2026.05		82.64	71.28	78.33
Llama-3.2-8B-Instruct-GRPO2026.05		82.18	75.21	78.73
ToolAlign2026.05		77.26	61.47	71.16
Toolformer2026.05		76.11	59.47	67.07
ReAct2026.05		73.58	58.43	66.08