IOI (Vals v2)
Code
Vals IOI v2 evaluates eighteen official IOI problems from 2024, 2025 and 2026 using OpenCode in an offline C++20 sandbox. Agents receive no submission or grading feedback during their attempts; only the final solution is graded. The score averages the three six-problem yearly scores. This protocol is explicitly incomparable with the earlier interactive-grading IOI board.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Argon | 100.0 |
| GPT-6 Astra | 100.0 |
| GPT-6.1 Sol | 96.9 |
| GPT-5.6 Sol | 91.2 |
| GPT-5.6 Terra | 87.6 |
| Claude Opus 5 | 84.3 |
| Claude Sonnet 5.5 | 83.1 |
| GPT-6 Sol | 82.6 |
| Kimi K3.1 Pro | 81.9 |
| Muse Wonder Full | 79.8 |
| Qwen3.8-Max | 68.9 |
| GLM-5.3 | 68.4 |
| Gemini 3.7 Flash | 67.8 |
| Muse Wonder | 66.9 |
| GPT-5.6 Luna | 61.8 |
| Hy4 Preview | 59.3 |
| Grok 4.7 | 57.7 |
| Gemini 3.8 Flash | 56.9 |
| Muse Spark 1.3 | 56.6 |
| GPT-6 Luna | 55.6 |
| GPT-5.3-Codex | 53.8 |
| GLM-5.3 Flash | 52.5 |
| Gemini 3.1 Pro Preview | 51.8 |
| DeepSeek V4 Pro 0813 | 51.6 |
| Kimi K3 | 48.9 |
| MiMo-V2.6-Flash | 47.7 |
| Grok 4.6 | 47.6 |
| Claude Haiku 5.5 | 47.3 |
| Ember-1 | 46.3 |
| Mistral Large 4 Preview | 45.3 |
| Sonnet 5 | 45.0 |
| Grok 4.5 | 40.6 |
| DeepSeek V4.1 Flash | 40.3 |
| MiMo-V2.6-Pro | 39.3 |
| Qwen3.8 27B | 39.1 |
| Step 5 Preview | 38.1 |
| Gemini 3.6 Flash | 35.1 |
| DeepSeek-V4-Flash-0731 | 32.7 |
| Beam 1.0 501B A23B | 22.8 |
| Muse Spark 1.2 | 21.8 |