VRP optimization-modeling on VRP Optimization-Modeling Benchmark 3
93.24Pass@1VRPCoder-GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| VRPCoder-GRPOModel scale=8B, Stage=GRPO2026.06 | 93.24 | |
| Gemini-3.1-Pro PreviewModel type=Closed-source frontier LLM2026.06 | 91.89 | |
| VRPCoder-SFTModel scale=8B, Stage=SFT2026.06 | 82.43 | |
| Claude-Sonnet-4.5Model type=Closed-source frontier LLM2026.06 | 79.73 | |
| DeepSeek-V3.2Model type=Open-source general-purpose LLM2026.06 | 51.35 | |
| MiniMax-M2.5Model type=Open-source general-purpose LLM2026.06 | 17.57 | |
| Qwen2.5-72B-InstructModel type=Open-source general-purpose LLM2026.06 | 12.16 | |
| ORLM-LLaMA-3-8BModel type=OR-LLM2026.06 | 12.16 | |
| OptMATH-Qwen2.5-7BModel type=OR-LLM2026.06 | 2.7 | |
| Qwen3-8BModel type=Open-source general-purpose LLM2026.06 | 1.35 | |
| SIRL-Gurobi-32BModel type=OR-LLM2026.06 | 1.35 | |
| SIRL-Gurobi-8BModel type=OR-LLM2026.06 | 0 |