Function Calling on Berkeley Function Calling Leaderboard (BFCL) Overall November 19, 2025
69.44Non-live AccuracyR2IF
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| R2IFModel=Llama3.2-3B-Instruct2026.04 | 69.44 | 56.23 | 90.95 | 72.21 | |
| ToolRLModel=Llama3.2-3B-Instruct2026.04 | 68.13 | 45 | 85.65 | 66.26 | |
| Binary RewardModel=Llama3.2-3B-Instruct2026.04 | 65.44 | 37.06 | 90.27 | 64.26 | |
| SFTModel=Llama3.2-3B-Instruct2026.04 | 60.19 | 33.45 | 75.61 | 56.41 | |
| R2IFModel=Qwen2.5-7B-Instruct2026.04 | 59.5 | 63.33 | 93.59 | 72.14 | |
| ToolRLModel=Qwen2.5-7B-Instruct2026.04 | 57.5 | 61.64 | 95.57 | 71.57 | |
| SFTModel=Qwen2.5-7B-Instruct2026.04 | 57.13 | 51.25 | 96 | 67.96 | |
| RawModel=Qwen2.5-7B-Instruct2026.04 | 55.5 | 56.07 | 87.85 | 66.47 | |
| Binary RewardModel=Qwen2.5-7B-Instruct2026.04 | 55.13 | 59.97 | 96.85 | 70.65 | |
| RawModel=Llama3.2-3B-Instruct2026.04 | 40.5 | 18.19 | 54.08 | 37.59 | |
| SFTModel=Qwen2.5-1.5B-Instruct2026.04 | 33.63 | 33.61 | 74.18 | 47.14 | |
| Binary RewardModel=Qwen2.5-1.5B-Instruct2026.04 | 33.25 | 38.41 | 98.51 | 56.72 | |
| R2IFModel=Qwen2.5-3B-Instruct2026.04 | 31.81 | 43.07 | 98.25 | 57.71 | |
| Binary RewardModel=Qwen2.5-3B-Instruct2026.04 | 31.69 | 39.9 | 98.62 | 56.74 | |
| ToolRLModel=Qwen2.5-1.5B-Instruct2026.04 | 31.19 | 35.8 | 95.15 | 54.04 | |
| SFTModel=Qwen2.5-3B-Instruct2026.04 | 31.19 | 40.46 | 83.64 | 51.76 | |
| RawModel=Qwen2.5-1.5B-Instruct2026.04 | 30.75 | 37.12 | 78.62 | 48.83 | |
| RawModel=Qwen2.5-3B-Instruct2026.04 | 30.5 | 35.09 | 92.44 | 52.68 | |
| R2IFModel=Qwen2.5-1.5B-Instruct2026.04 | 29.38 | 37.62 | 96.17 | 54.39 | |
| ToolRLModel=Qwen2.5-3B-Instruct2026.04 | 28 | 27.68 | 98.64 | 51.44 |