IOI 2025 (Vals v2)
Code
The six IOI 2025 problems under Vals' v2 OpenCode protocol without interactive grading feedback. Final solutions receive official subtask points, averaged across the problems.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 100.0 |
| GPT-6 Astra | 100.0 |
| GPT-5.6 Sol | 88.3 |
| Claude Fable 5.1 | 87.2 |
| Claude Opus 5 | 87.2 |
| GPT-6 Sol | 79.3 |
| GPT-5.6 Terra | 77.8 |
| GLM-5.3 | 75.5 |
| Gemini 3.7 Flash | 69.8 |
| Hy4 Preview | 68.5 |
| Qwen3.8-Max | 68.3 |
| Gemini 3.8 Flash | 61.3 |
| Muse Spark 1.3 | 61.2 |
| GPT-5.3-Codex | 57.0 |
| GLM-5.3 Flash | 56.7 |
| Grok 4.7 | 56.2 |
| GPT-5.6 Luna | 49.8 |
| Grok 4.6 | 49.0 |
| DeepSeek V4 Pro 0813 | 48.5 |
| Qwen3.8 27B | 48.5 |
| Sonnet 5 | 42.0 |
| Gemini 3.6 Flash | 40.8 |
| MiMo-V2.6-Pro | 39.3 |
| GPT-6 Luna | 39.0 |
| DeepSeek V4.1 Flash | 35.2 |
| Grok 4.5 | 33.5 |
| Kimi K3 | 29.3 |
| DeepSeek-V4-Flash-0731 | 28.2 |
| Muse Spark 1.2 | 23.2 |
| Inkling | 21.0 |
| Inkling-Small | 9.2 |
| Mercury 2.5 | 3.5 |