Code Generation on OJBench
58.8AccuracyGemini 3 Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 3 ProParams=N/A2026.06 | 58.8 | |
| GLM-5Params=744B2026.06 | 55 | |
| Kimi K2.5Params=1T2026.06 | 54.7 | |
| DeepSeek V3.2Params=671B2026.06 | 48.4 | |
| GPT-OSS (high)Params=120B2026.06 | 41.5 | |
| Gemini 2.5 ProParams=N/A2026.06 | 38.9 | |
| VibeThinker-3BParams=3B2026.06 | 38.6 | |
| DeepSeek R1 0528Params=671B2026.06 | 33.6 | |
| OpenAI o3 (high)Params=N/A2026.06 | 25.4 | |
| Gemini 2.5 FlashParams=N/A2026.06 | 23.5 |