IOI
Competitive-programming ability on International Olympiad in Informatics problems, with the agent seeing only what a contestant sees: the statement, the sample grader and one sample.
How it is run
Three contest years (2024, 2025 and 2026) run inside an OpenCode agent in an isolated sandbox with no tests, no internet and no submission feedback, so every point comes from code the model wrote and checked itself. Vals also reports accuracy per year cohort.
What it cannot tell you
Contest-year breakdowns matter: a model can perform differently on the 2024, 2025 and 2026 cohorts. Problems published after a training cut-off can still be solved, so recency alone does not establish difficulty.
- Leader
- Gemini 4 Argon100.00% · Vals AI's own list
- Top-ten spread
- 17.39%Leader minus tenth place, or last place in smaller fields.
- Median
- 54.70%of the 38 rows published
- No published result
- 15of 46 models on this site
Independent leaderboard · Vals AI
Top 38 of 38 runs by Vals AI, one harness.
| # | Model | Score |
|---|---|---|
| 1 | Gemini 4 Argon | 100.00% |
| 1 | GPT-6 Astra | 100.00% |
| 3 | GPT-6.1 Sol | 96.89% |
| 4 | Claude Opus 5.5 | 95.06% |
| 5 | GPT-5.6 Sol | 91.17% |
| 6 | Claude Fable 5.1 | 90.78% |
| 7 | GPT-5.6 Terra | 87.61% |
| 8 | Claude Opus 5 | 84.33% |
| 9 | Claude Sonnet 5.5 | 83.06% |
| 10 | GPT-6 Sol | 82.61% |
| 11 | Qwen3.8 Max | 68.89% |
| 12 | GLM 5.3 | 68.44% |
| 13 | Gemini 3.7 Flash | 67.83% |
| 14 | GPT-5.6 Luna | 61.78% |
| 15 | Hunyuan Hy4 Preview | 59.33% |
| 16 | Grok 4.7 | 57.72% |
| 17 | Gemini 3.8 Flash | 56.94% |
| 18 | Muse Spark 1.3 Max | 56.56% |
| 19 | GPT-6 Luna | 55.56% |
| 20 | Gpt 5.3 Codex | 53.83% |
| 21 | GLM 5.3 Flash | 52.50% |
| 22 | Gemini 3.1 Pro | 51.83% |
| 23 | DeepSeek V4 Pro | 51.61% |
| 24 | Kimi K3 | 48.94% |
| 25 | MiMo-V2.6-Flash | 47.67% |
| 26 | Grok 4.6 | 47.61% |
| 27 | Claude Sonnet 5 | 45.00% |
| 28 | Muse Spark 1 3 | 43.94% |
| 29 | Grok 4.5 | 40.56% |
| 30 | DeepSeek V4.1 Flash | 40.28% |
| 31 | MiMo V2.6 Pro | 39.33% |
| 32 | Qwen3.8 27b | 39.06% |
| 33 | Gemini 3.6 Flash | 35.06% |
| 34 | Deepseek V4 Flash 0731 | 32.72% |
| 35 | Muse Spark 1.2 | 21.78% |
| 36 | Inkling | 14.94% |
| 37 | Inkling Small | 9.33% |
| 38 | Mercury 2.5 | 2.44% |
No published result tracked
15 of the 46 models on this site have no published number on this benchmark, so their cells in the home table stay blank rather than reading as zero.
Read this with
Benchmarks in the same category, which usually disagree with this one on purpose.
Sources
Every page used for this benchmark, read on 17 October 2026.
- Benchmark site or the page that publishes itioi2024.org/
- Independent resultswww.vals.ai/benchmarks/ioi