LiveOIBench — leaderboard
Contamination-free Olympiad benchmark with 403 expert-curated problems from 14 Informatics Olympiads (IOI, USACO, APIO, etc). Models ranked by human percentile against actual contestants.
Metric: Avg Human Percentile. Source: liveoibench.github.io. Status: saturation imminent. 58 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 80.18 |
| 2 | GPT-OSS-120B (High) | 71.85 |
| 3 | Grok 4 Fast (Reasoning) | 69.8 |
| 4 | Gemini 2.5 Pro | 67.44 |
| 5 | O3 Mini (High) | 60.86 |
| 6 | GPT-OSS-120B (Medium) | 57.55 |
| 7 | GPT-OSS-20B (High) | 57.13 |
| 8 | Gemini 2.5 Flash | 53.75 |
| 9 | Claude Sonnet 4.5 | 51.39 |
| 10 | GPT-OSS-20B (Medium) | 51.12 |
| 11 | GPT-OSS-120B (Low) | 44.07 |
| 12 | Qwen 3 32B | 41.95 |
| 13 | DeepSeek R1 | 38.8 |
| 14 | Qwen 3 30B A3B | 35.2 |
| 15 | GPT-OSS-20B (Low) | 34.64 |
Interactive version: theaggregate.ai/benchmark?slug=liveoibench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.