Function Calling on BFCL V3
79.3Overall AccuracyD-CORE-14B
Evaluation Results
| Method | Links | |||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| D-CORE-14BModel Category=Custom-Trained Model, Backbone=Qwen3-14B, Parameters=14B2026.02 | 79.3 | — | — | 82.9 | 67.4 | 89.2 | 72.2 | — | — | — | — | 87.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HardGen-4B-RLAccess=Open-source, Parameter Counts=4B, Training=Reinforcement Learning2026.01 | 79.14 | — | — | — | — | — | — | 68.5 | 64.5 | 50.5 | 69 | — | — | — | 90.73 | 83.55 | 87.14 | 63.13 | — | — | — | — | — | — | — | |
| Claude-Opus-4-5-20251101Access=Closed-source, Parameter Counts=-2026.01 | 78.92 | — | — | — | — | — | — | 81 | 64 | 58 | 70.5 | — | — | — | 88.58 | 79.79 | 84.19 | 68.38 | — | — | — | — | — | — | — | |
| GLM-4.6Access=Open-source, Parameter Counts=355B2026.01 | 78.82 | — | — | — | — | — | — | 74.5 | 68 | 63 | 66.5 | — | — | — | 87.56 | 80.9 | 84.23 | 68 | — | — | — | — | — | — | — | |
| xLAM2-70BModel Category=Open-Source Model, Parameters=70B2026.02 | 78.4 | — | — | 72.9 | 75 | 78.9 | 66.7 | — | — | — | — | 88.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gemini-3-Pro-PreviewAccess=Closed-source, Parameter Counts=-2026.01 | 78.17 | — | — | — | — | — | — | 64.5 | 60 | 54.5 | 64 | — | — | — | 90.65 | 83.12 | 86.89 | 60.75 | — | — | — | — | — | — | — | |
| D-CORE-8BModel Category=Custom-Trained Model, Backbone=Qwen3-8B, Parameters=8B2026.02 | 77.7 | — | — | 82.2 | 63.8 | 88.2 | 77.8 | — | — | — | — | 86.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Claude-Sonnet-4-5-20250929Access=Closed-source, Parameter Counts=-2026.01 | 77.05 | — | — | — | — | — | — | 69 | 65 | 52.5 | 59 | — | — | — | 88.65 | 81.13 | 84.89 | 61.38 | — | — | — | — | — | — | — | |
| xLAM2-32BModel Category=Open-Source Model, Parameters=32B2026.02 | 76.4 | — | — | 74.2 | 67.1 | 76.7 | 88.9 | — | — | — | — | 89.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Grok-4-1-fast-reasoningAccess=Closed-source, Parameter Counts=-2026.01 | 75.2 | — | — | — | — | — | — | 70.5 | 59.5 | 43 | 62.5 | — | — | — | 88.27 | 78.46 | 83.37 | 58.88 | — | — | — | — | — | — | — | |
| WATT-TOOL-70BScale=70B, Protocol=FC2025.03 | 74.31 | 84.06 | 89.39 | 77.74 | 58.75 | 94.44 | 76.32 | 67.5 | 57.5 | 48.5 | 61.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HardGen-4B-SFTAccess=Open-source, Parameter Counts=4B, Training=Supervised Fine-Tuning2026.01 | 73.7 | — | — | — | — | — | — | 55.2 | 49.1 | 41.4 | 52.9 | — | — | — | 89.03 | 82.42 | 85.73 | 49.65 | — | — | — | — | — | — | — | |
| Claude-Haiku-4-5-20251001Access=Closed-source, Parameter Counts=-2026.01 | 72.93 | — | — | — | — | — | — | 63.5 | 42.5 | 52.5 | 56 | — | — | — | 86.5 | 78.68 | 82.59 | 53.63 | — | — | — | — | — | — | — | |
| Qwen3-32BAccess=Open-source, Parameter Counts=32B2026.01 | 72.88 | — | — | — | — | — | — | 56 | 52.5 | 40 | 43 | — | — | — | 88.77 | 82.01 | 85.39 | 47.88 | — | — | — | — | — | — | — | |
| GPT-4o-2024-11-20Protocol=PROMPT2025.03 | 72.08 | 88.1 | 89.38 | 79.83 | 47.62 | 83.33 | 83.76 | 59 | 41 | 35.5 | 55 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| xLAM2-8BModel Category=Open-Source Model, Parameters=8B2026.02 | 72 | — | — | 66.7 | 69.1 | 64.3 | 77.8 | — | — | — | — | 84.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4oModel Category=Proprietary Model2026.02 | 71.7 | — | — | 78.8 | 50 | 81.3 | 83.3 | — | — | — | — | 86.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-235B-A22B-InstructAccess=Open-source, Parameter Counts=235B2026.01 | 71.21 | — | — | — | — | — | — | 54 | 42.5 | 31.5 | 50.5 | — | — | — | 90.33 | 78.68 | 84.51 | 44.63 | — | — | — | — | — | — | — | |
| Grok-4-1-fast-non-reasoningAccess=Closed-source, Parameter Counts=-2026.01 | 70.94 | — | — | — | — | — | — | 58 | 39.5 | 37.5 | 52 | — | — | — | 88.13 | 77.94 | 83.04 | 46.75 | — | — | — | — | — | — | — | |
| Nanbeige4-3B-Thinking-2511Access=Open-source, Parameter Counts=3B2026.01 | 70.71 | — | — | — | — | — | — | 58.5 | 54 | 45 | 47 | — | — | — | 81.58 | 79.42 | 80.5 | 51.12 | — | — | — | — | — | — | — | |
| Kimi-K2-InstructAccess=Open-source, Parameter Counts=1043B2026.01 | 70.3 | — | — | — | — | — | — | 62 | 41 | 44.5 | 55 | — | — | — | 81.6 | 78.68 | 80.14 | 50.63 | — | — | — | — | — | — | — | |
| AWMModel Size=14B2026.02 | 70.18 | — | 81.46 | 77.2 | 51.88 | 78.37 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4o-2024-11-20Protocol=FC2025.03 | 69.58 | 87.42 | 89.2 | 79.65 | 41 | 83.33 | 83.15 | 62.5 | 6 | 37.5 | 58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-32BModel Category=Open-Source Model, Parameters=32B2026.02 | 69.2 | — | — | 77.3 | 43.1 | 75.8 | 72.2 | — | — | — | — | 88.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepSeek-V3.2-ExpAccess=Open-source, Parameter Counts=671B2026.01 | 68.81 | — | — | — | — | — | — | 55 | 49 | 27 | 48.5 | — | — | — | 85.52 | 76.02 | 80.77 | 44.88 | — | — | — | — | — | — | — | |
| ToolRL-Qwen3-14BModel Category=Custom-Trained Model, Backbone=Qwen3-14B2026.02 | 68.5 | — | — | 81.5 | 35.8 | 81.9 | 72.2 | — | — | — | — | 89.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| xLAM-2-3b-fc-rAccess=Open-source, Parameter Counts=3B2026.01 | 68.09 | — | — | — | — | — | — | 71.5 | 59 | 57.5 | 45.5 | — | — | — | 82.96 | 62.92 | 72.94 | 58.38 | — | — | — | — | — | — | — | |
| MAGNET-14B-mDPOScale=14B, Strategy=mDPO2025.03 | 68.01 | 90.13 | 89.75 | 79.14 | 37.88 | 88.89 | 84.78 | 52 | 36 | 35.5 | 28 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4-TURBO-2024-04-092025.03 | 67.88 | 84.73 | 85.21 | 80.5 | 38.12 | 72.22 | 83.81 | 54 | 13.5 | 35.5 | 49.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| o1Model Category=Proprietary Model2026.02 | 67.8 | — | — | 80.6 | 36.1 | 87.8 | 72.2 | — | — | — | — | 85.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Watt-ToolParameters=8B2026.02 | 67.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TeacherParameters=8B2026.02 | 67.74 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimulatorModel Size=14B2026.02 | 67.68 | — | 77.23 | 73.43 | 52.38 | 76.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolACE-2-8BAccess=Open-source, Parameter Counts=8B2026.01 | 67.63 | — | — | — | — | — | — | 49 | 28 | 30.5 | 46 | — | — | — | 87.1 | 77.42 | 82.26 | 38.38 | — | — | — | — | — | — | — | |
| WATT-TOOL-8B*Scale=8B, Protocol=FC, reproduced=true2025.03 | 67.33 | 86.44 | 87.73 | 76.23 | 38.25 | 77.78 | 82.89 | 46 | 40 | 27 | 40 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGNET-14B-SFTScale=14B, Strategy=SFT2025.03 | 66.83 | 90.02 | 88.2 | 77.92 | 33.38 | 72.22 | 82.59 | 47 | 32 | 32 | 22.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| o1-2024-12-17Protocol=PROMPT2025.03 | 66.73 | 78.92 | 82.7 | 78.14 | 28.25 | 61.11 | 89.62 | 40.5 | 5 | 34.5 | 33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3Parameters=8B2026.02 | 66.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-8BModel Category=Open-Source Model, Parameters=8B2026.02 | 66.3 | — | — | 78.5 | 33 | 79.1 | 77.8 | — | — | — | — | 88.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AWMModel Size=8B2026.02 | 65.94 | — | 80.44 | 72.39 | 45 | 70.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-14BModel Category=Open-Source Model, Parameters=14B2026.02 | 65.9 | — | — | 80 | 36.6 | 81 | 72.2 | — | — | — | — | 88.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRL-Qwen3-8BModel Category=Custom-Trained Model, Backbone=Qwen3-8B2026.02 | 65.9 | — | — | 82.4 | 26.8 | 85.9 | 77.8 | — | — | — | — | 88.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4o-2024-11-20Access=Closed-source, Parameter Counts=-2026.01 | 65.64 | — | — | — | — | — | — | 55.5 | 34.5 | 29 | 51 | — | — | — | 83.88 | 70.54 | 77.21 | 42.5 | — | — | — | — | — | — | — | |
| ToolACE-MTAccess=Open-source, Parameter Counts=8B2026.01 | 65.57 | — | — | — | — | — | — | 57.5 | 31.5 | 34 | 38 | — | — | — | 84.94 | 71.52 | 78.23 | 40.25 | — | — | — | — | — | — | — | |
| STARParameters=4B2026.02 | 65.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gemini-2.5-FlashAccess=Closed-source, Parameter Counts=-2026.01 | 65.2 | — | — | — | — | — | — | 41.5 | 36 | 32 | 35.5 | — | — | — | 84.96 | 74.39 | 79.68 | 36.25 | — | — | — | — | — | — | — | |
| MAGNET-7B-mDPOScale=7B, Strategy=mDPO2025.03 | 64.64 | 89.4 | 89.27 | 77.92 | 27.75 | 83.33 | 78.51 | 39 | 24 | 26 | 22 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-30B-A3B-ThinkingAccess=Open-source, Parameter Counts=30B2026.01 | 64.57 | — | — | — | — | — | — | 43.5 | 10.5 | 25 | 41 | — | — | — | 85.77 | 77.94 | 81.86 | 30 | — | — | — | — | — | — | — | |
| AWMModel Size=4B2026.02 | 64.5 | — | 78.6 | 76.91 | 39.38 | 73.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepSeek-R1Model Category=Proprietary Model2026.02 | 63.8 | — | — | 77.3 | 38.9 | 73.3 | 77.8 | — | — | — | — | 75.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3Parameters=4B2026.02 | 63.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGNET-7B-SFTScale=7B, Strategy=SFT2025.03 | 62.73 | 88.6 | 85.73 | 74.19 | 26.5 | 66.67 | 78.67 | 35.5 | 24 | 27.5 | 19 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRL+DiPOBackbone=Qwen2.5-7B-Instruct2026.04 | 62.51 | — | — | 76.83 | 24.5 | 87.5 | 69.57 | — | — | — | — | 86.21 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Hammer2.1Parameters=7B2026.02 | 62.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gemini-1.5-Pro-002Protocol=Prompt2025.03 | 62.19 | 88.58 | 91.27 | 76.72 | 20.75 | 72.22 | 78.15 | 23 | 19.5 | 17.5 | 23 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-4B (Base Model)Access=Open-source, Parameter Counts=4B2026.01 | 62.13 | — | — | — | — | — | — | 26.5 | 21 | 15.5 | 25.5 | — | — | — | 87.88 | 76.39 | 82.14 | 22.13 | — | — | — | — | — | — | — | |
| Hammer2.1-7bScale=7B, Protocol=FC2025.03 | 61.83 | 88.65 | 85.48 | 75.11 | 23.5 | 82.35 | 78.59 | 35.5 | 25.5 | 19 | 14 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gemini-2.0-Flash-ExpProtocol=Prompt2025.03 | 61.74 | 89.96 | 79.89 | 82.01 | 17.88 | 77.78 | 86.44 | 28 | 3 | 19 | 21.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaseModel Size=14B2026.02 | 61.25 | — | 69.48 | 67.28 | 47 | 77.94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRL+DAPOBackbone=Qwen2.5-7B-Instruct2026.04 | 61.06 | — | — | 76.31 | 19.75 | 87.5 | 67.25 | — | — | — | — | 87.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Llama-4-MaverickAccess=Open-source, Parameter Counts=400B2026.01 | 60.85 | — | — | — | — | — | — | 27 | 22 | 14 | 18 | — | — | — | 88.65 | 73.65 | 81.15 | 20.25 | — | — | — | — | — | — | — | |
| GPT-5.2-2025-12-11Access=Closed-source, Parameter Counts=-2026.01 | 60.12 | — | — | — | — | — | — | 36.5 | 18 | 27.5 | 30.5 | — | — | — | 81.85 | 70.39 | 76.12 | 28.13 | — | — | — | — | — | — | — | |
| Hammer2.1Parameters=3B2026.02 | 59.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Claude-3.7-SonnetModel Category=Proprietary Model2026.02 | 58.6 | — | — | 78.4 | 48.4 | 81.4 | 72.2 | — | — | — | — | 41.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRL+GRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 58.38 | — | — | 74.9 | 18.12 | 83.33 | 76.68 | — | — | — | — | 86.17 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Functionary-Small-v3.1Scale=7B, Protocol=FC2025.03 | 56.49 | 86.75 | 87.12 | 73.75 | 10.12 | 77.78 | 70.89 | 18 | 2.5 | 14 | 6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STARParameters=1.7B2026.02 | 56.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimulatorModel Size=4B2026.02 | 55.52 | — | 66.1 | 62.69 | 37.75 | 75.48 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRL+DiPOBackbone=Qwen2.5-3B-Instruct2026.04 | 55.03 | — | — | 73.06 | 8.62 | 100 | 54.16 | — | — | — | — | 83.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaseModel Size=4B2026.02 | 54.92 | — | 61.44 | 63.95 | 39.38 | 73.93 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3Parameters=1.7B2026.02 | 54.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EnvScalerModel Size=4B2026.02 | 54.06 | — | 60.31 | 64.25 | 37.63 | 85.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaseModel Size=8B2026.02 | 53.83 | — | 59.58 | 58.03 | 43.88 | 76.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRL+DAPOBackbone=Qwen2.5-3B-Instruct2026.04 | 53.21 | — | — | 69.43 | 8 | 81.25 | 57.6 | — | — | — | — | 82.19 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-Coder-7B-InstructScale=7B2025.03 | 53.13 | 86.83 | 82.27 | 66.99 | 8.25 | 88.89 | 65.39 | 11.5 | 6.5 | 5.5 | 5.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRL+GRPOBackbone=Qwen2.5-3B-Instruct2026.04 | 52.98 | — | — | 73.78 | 3.75 | 100 | 56.44 | — | — | — | — | 81.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimulatorModel Size=8B2026.02 | 52.53 | — | 56.35 | 59.36 | 41.88 | 74.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-Coder-14B-InstructScale=14B2025.03 | 51.88 | 90.94 | 87.8 | 65.3 | 5.38 | 100 | 44.58 | 7.5 | 7 | 4 | 3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| STARParameters=0.6B2026.02 | 51.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CKD+Sim-RLBackbone=Qwen3-0.6B, Framework=STAR2026.02 | 51.7 | — | — | 68.19 | 7 | — | — | — | — | — | — | 78.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPO, Cold StartBackbone=Qwen2.5-3B-Instruct2026.04 | 51.15 | — | — | 67.78 | 4.88 | 100 | 18.09 | — | — | — | — | 82.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT+Sim-RLBackbone=Qwen3-0.6B, Note=Sim-RL on SFT-think2026.02 | 50.41 | — | — | 66.99 | 6.13 | — | — | — | — | — | — | 76.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CKDBackbone=Qwen3-0.6B2026.02 | 49.84 | — | — | 66.15 | 5.62 | — | — | — | — | — | — | 75.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Llama3.1Parameters=8B2026.02 | 49.57 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FKLBackbone=Qwen3-0.6B2026.02 | 49.51 | — | — | 65.93 | 5.12 | — | — | — | — | — | — | 76.44 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Sim-RLBackbone=Qwen3-0.6B2026.02 | 49.35 | — | — | 67.39 | 3.25 | — | — | — | — | — | — | 75.21 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LUFFYBackbone=Qwen3-0.6B2026.02 | 49.23 | — | — | 64.59 | 5.48 | — | — | — | — | — | — | 76.75 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT-thinkBackbone=Qwen3-0.6B2026.02 | 47.59 | — | — | 64.46 | 4.5 | — | — | — | — | — | — | 71.54 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ToolRLBackbone=Qwen3-0.6B2026.02 | 47.35 | — | — | 66.55 | 6.75 | — | — | — | — | — | — | 64.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3Parameters=0.6B2026.02 | 47.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Base-modelBackbone=Qwen3-0.6B2026.02 | 47.33 | — | — | 65.66 | 1.88 | — | — | — | — | — | — | 71.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GKDBackbone=Qwen3-0.6B2026.02 | 47.32 | — | — | 67.61 | 3.25 | — | — | — | — | — | — | 67.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPO, Cold StartBackbone=Qwen2.5-7B-Instruct2026.04 | 46.68 | — | — | 63.17 | 0.38 | 88.89 | 52.92 | — | — | — | — | 79.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT400+GRPOBackbone=Qwen2.5-3B-Instruct2026.04 | 46.42 | — | — | 64.15 | 1.75 | 94.44 | 58.63 | — | — | — | — | 76.21 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Llama3.2Parameters=3B2026.02 | 45.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT400+PPOBackbone=Qwen2.5-3B-Instruct2026.04 | 45.8 | — | — | 58.76 | 5.12 | 100 | 48.4 | — | — | — | — | 78.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFTBackbone=Qwen3-0.6B2026.02 | 44.58 | — | — | 62.15 | 1.62 | — | — | — | — | — | — | 66.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT400+PPOBackbone=Qwen2.5-7B-Instruct2026.04 | 42.02 | — | — | 51.84 | 0.25 | 100 | 29.66 | — | — | — | — | 83.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Base modelBackbone=Qwen2.5-7B-Instruct2026.04 | 41.97 | — | — | 53.51 | 4.25 | 76.47 | 62.66 | — | — | — | — | 66.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SFT400+GRPOBackbone=Qwen2.5-7B-Instruct2026.04 | 39.25 | — | — | 46.51 | 0.25 | 100 | 14.19 | — | — | — | — | 80.69 | — | — | — | — | — | — | — | — | — | — | — | — | — |