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 |
|---|---|---|
| Q3 2024 | 5% | Gemini-1.5-Pro-002 |
| Q4 2024 | 8% | o1 (unknown thinking) |
| Q1 2025 | 18% | Gemini 2.5 Pro Exp (Mar 2025) |
| Q2 2025 | 22% | Gemini 2.5 Pro Preview (Jun 2025) |
| Q3 2025 | 25% | GPT-5 (high) |
| Q4 2025 | 38% | Gemini 3 Pro Preview |
| Q1 2026 | 46% | Gemini 3.1 Pro Preview |
| Q2 2026 | 46% | Gemini 3.1 Pro Preview |
| Q3 2026 | 55% | GPT-6 Astra (unknown thinking) |
| Q4 2026 | 55% | GPT-6 Astra (unknown thinking) |
Room left to improve
Closing in
Best score 55% out of 100%.
Models get more than half right and are improving quickly. Watch this one over the next year.
Humans on this test
There is no overall human score. Each question was written by a specialist to be hard for other specialists, and no one person could answer across every field.
What it tests
About 2,500 questions across more than 100 subjects, written by around a thousand subject experts and kept only if leading models of the time got them wrong.
Why it matters
It was designed as a broad academic exam that AI would struggle with for years.
Caveats
Because questions were chosen to beat older models, early scores start artificially near zero. Results come from the official leaderboard rather than Epoch AI’s own runs.
How benchmarks work →Model scores
Top 10 of 48 models by score
| Rank | Model | Developer | Score | Released |
|---|---|---|---|---|
| 1 | GPT-6 Astra (unknown thinking) OpenAI | OpenAI | 55% | Sep 2026 |
| 2 | Claude Fable 5.1 (xhigh) Anthropic | Anthropic | 47% | Sep 2026 |
| 3 | Gemini 3.1 Pro Preview Google DeepMind | Google DeepMind | 46% | Feb 2026 |
| 4 | Gemini 3.8 Flash (unknown) Google DeepMind | Google DeepMind | 45% | Sep 2026 |
| 5 | GPT-5.4 Pro OpenAI | OpenAI | 44% | Mar 2026 |
| 6 | Muse Spark Meta AI | Meta AI | 41% | Apr 2026 |
| 7 | Gemini 3 Pro Preview Google DeepMind | Google DeepMind | 38% | Nov 2025 |
| 8 | GPT-5.4 (xhigh) OpenAI | OpenAI | 36% | Mar 2026 |
| 9 | Claude Opus 4.7 (unknown) Anthropic | Anthropic | 36% | Apr 2026 |
| 10 | Claude Opus 4.6 (max) Anthropic | Anthropic | 34% | Feb 2026 |
Benchmark data from Epoch AI (CC BY 4.0) and official leaderboards. Updated weekly, last on . Download CSV Methodology