Benchmarks / Coding
IOI
Vals AI's overall score for version 2 of the International Olympiad in Informatics benchmark.
IOI
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Gemini 4 | high | 100.0% |
| 2 | GPT-6 Astra | max | 100.0% |
| 3 | GPT-6.1 Sol | max | 96.9% |
| 4 | Opus 5.5 | max | 95.1% |
| 5 | GPT-5.6 Sol | max | 91.2% |
| 6 | Fable 5.1 | max | 90.8% |
| 7 | GPT-5.6 Terra | max | 87.6% |
| 8 | Opus 5 | max | 84.3% |
| 9 | Sonnet 5.5 | max | 83.1% |
| 10 | GPT-6 Sol | max | 82.6% |
| 11 | Wonder Full | max | 79.8% |
| 12 | Qwen 3.8 Max | max | 68.9% |
| 13 | GLM 5.3 | max | 68.4% |
| 14 | Gemini 3.7 Flash | high | 67.8% |
| 15 | Wonder | max | 66.9% |
| 16 | GPT-5.6 Luna | max | 61.8% |
| 17 | Hy4 | – | 59.3% |
| 18 | Grok 4.7 | xhigh | 57.7% |
| 19 | Gemini 3.8 Flash | high | 56.9% |
| 20 | Muse Spark 1.3 | max | 56.6% |
| 21 | GPT-6 Luna | max | 55.6% |
| 22 | GPT-5.3 Codex | xhigh | 53.8% |
| 23 | GLM 5.3 Flash | max | 52.5% |
| 24 | Gemini 3.1 Pro | high | 51.8% |
| 25 | DeepSeek V4 Pro 0813 | max | 51.6% |
| 26 | Kimi K3 | max | 48.9% |
| 27 | MiMo V2.6 Flash | high | 47.7% |
| 28 | Grok 4.6 | high | 47.6% |
| 29 | Haiku 5.5 | max | 47.3% |
| 30 | Ember-1 | – | 46.3% |
| 31 | Mistral Large 4 | high | 45.3% |
| 32 | Sonnet 5 | max | 45.0% |
| 33 | Grok 4.5 | high | 40.6% |
| 34 | DeepSeek V4.1 Flash | high | 40.3% |
| 35 | MiMo V2.6 Pro | high | 39.3% |
| 36 | Qwen 3.8 27B | xhigh | 39.1% |
| 37 | Step 5 Preview | – | 38.1% |
| 38 | Gemini 3.6 Flash | high | 35.1% |
| 39 | DeepSeek V4 Flash 0731 | high | 32.7% |
| 40 | Muse Spark 1.2 | xhigh | 21.8% |
| 41 | Inkling | max | 14.9% |
| 42 | Inkling Small | xhigh | 9.3% |
| 43 | Mercury 2.5 | high | 2.4% |