How good is AI getting?
Pick a test. See the best AI score over time, what the test actually measures, and how it compares with people.
Hover or tap a point to see that quarter. The line shows the best score any model had reached by then.
| Quarter | Best score | Record held by |
|---|---|---|
| Q1 2024 | 0% | GPT-3.5 Turbo (Jan 2024) |
| Q2 2024 | 1% | gpt-4-turbo-2024-04-09 |
| Q3 2024 | 1% | gpt-4-turbo-2024-04-09 |
| Q4 2024 | 15% | o1 (high) |
| Q1 2025 | 19% | o3-mini (high) |
| Q2 2025 | 36% | o4-mini (high) |
| Q3 2025 | 55% | GPT-5 (high) |
| Q4 2025 | 74% | GPT-5.2 Pro |
| Q1 2026 | 83% | GPT-5.4 Pro (xhigh) |
| Q2 2026 | 88% | GPT-5.5 Pro (xhigh) |
| Q3 2026 | 94% | GPT-6.1 Sol (max) |
| Q4 2026 | 94% | GPT-6.1 Sol (max) |
Room left to improve
Nearly maxed out
Best score 94% out of 100%.
Models already answer most questions correctly. This test is close to its ceiling and will soon stop telling us much.
Humans on this test
No single human baseline exists. The problems were written by research mathematicians, and many take a specialist hours or days. No group of people has sat the whole test.
What it tests
Hundreds of unpublished problems written by research mathematicians, with answers that can be checked automatically. This site tracks tiers 1 to 3; a harder fourth tier is scored separately.
Why it matters
The problems are unpublished, so progress signals mathematical reasoning rather than memory.
Caveats
The problems are not public, but OpenAI commissioned them and has access to most of the problems and their solutions. Epoch AI, which runs the test, keeps a set back from OpenAI to check results against. Scores here are for the current version of the set, re-run on older models.
How benchmarks work →Model scores
Top 10 of 50 models by score
| Rank | Model | Developer | Score | Released |
|---|---|---|---|---|
| 1 | GPT-6 Astra (max) OpenAI | OpenAI | 94% | Sep 2026 |
| 2 | GPT-6.1 Sol (max) OpenAI | OpenAI | 94% | Sep 2026 |
| 3 | Claude Opus 5.5 (max) Anthropic | Anthropic | 91% | Sep 2026 |
| 4 | Claude Fable 5.1 (max) Anthropic | Anthropic | 90% | Sep 2026 |
| 5 | GPT-6 Sol (max) OpenAI | OpenAI | 90% | Sep 2026 |
| 6 | GPT-5.6 Sol (max) OpenAI | OpenAI | 89% | Jul 2026 |
| 7 | Claude Sonnet 5.5 (max) Anthropic | Anthropic | 89% | Sep 2026 |
| 8 | GPT-5.5 Pro (xhigh) OpenAI | OpenAI | 88% | Apr 2026 |
| 9 | Claude Fable 5 (max) Anthropic | Anthropic | 87% | Jun 2026 |
| 10 | GPT-5.6 Terra (max) OpenAI | OpenAI | 86% | Jul 2026 |
Benchmark data from Epoch AI (CC BY 4.0) and official leaderboards. Updated weekly, last on . Download CSV Methodology