OJBench (C++)
Code · 2025-06-19
OJBench's 232 NOI and ICPC competition problems solved in C++, judged by hidden test cases. Scores report pass@1, the percentage of problems whose generated C++ program passes every test. Higher is better.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3 Pro Preview | 68.5 |
| Kimi K2.5 | 57.4 |
| GPT-5 | 56.2 |
| DeepSeek-V3.2 | 54.7 |
| Opus 4.5 | 54.6 |
| Kimi K2 Thinking | 48.7 |
| DeepSeek-V3.2-Exp | 38.2 |
| Sonnet 4.5 | 30.4 |
| Kimi K2 Instruct 0905 | 25.5 |