VRP Optimization-Modeling on VRP Optimization-Modeling Benchmark 2
96.97Pass@1VRPCoder-GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| VRPCoder-GRPOModel scale=8B, Stage=GRPO2026.06 | 96.97 | |
| Gemini-3.1-Pro PreviewModel type=Closed-source frontier LLM2026.06 | 89.39 | |
| VRPCoder-SFTModel scale=8B, Stage=SFT2026.06 | 89.39 | |
| Claude-Sonnet-4.5Model type=Closed-source frontier LLM2026.06 | 42.42 | |
| MiniMax-M2.5Model type=Open-source general-purpose LLM2026.06 | 9.09 | |
| DeepSeek-V3.2Model type=Open-source general-purpose LLM2026.06 | 7.58 | |
| Qwen2.5-72B-InstructModel type=Open-source general-purpose LLM2026.06 | 6.06 | |
| SIRL-Gurobi-32BModel type=OR-LLM2026.06 | 3.03 | |
| Qwen3-8BModel type=Open-source general-purpose LLM2026.06 | 0 | |
| ORLM-LLaMA-3-8BModel type=OR-LLM2026.06 | 0 | |
| OptMATH-Qwen2.5-7BModel type=OR-LLM2026.06 | 0 | |
| SIRL-Gurobi-8BModel type=OR-LLM2026.06 | 0 |