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 2023 | 36% | GPT-4 (Mar 2023) |
| Q2 2023 | 36% | GPT-4 (Mar 2023) |
| Q3 2023 | 36% | GPT-4 (Mar 2023) |
| Q4 2023 | 42% | GPT-4 Turbo Preview (Nov 2023) |
| Q1 2024 | 47% | claude-3-opus-20240229 |
| Q2 2024 | 54% | Claude 3.5 Sonnet (Jun 2024) |
| Q3 2024 | 62% | o1-mini (high) |
| Q4 2024 | 77% | o1 (high) |
| Q1 2025 | 84% | Gemini 2.5 Pro Exp (Mar 2025) |
| Q2 2025 | 85% | gemini-2.5-pro |
| Q3 2025 | 87% | grok-4-0709 |
| Q4 2025 | 93% | Gemini 3 Pro Preview |
| Q1 2026 | 95% | GPT-5.4 Pro (xhigh) |
| Q2 2026 | 95% | GPT-5.4 Pro (xhigh) |
| Q3 2026 | 96% | GPT-6 Astra (max) |
| Q4 2026 | 96% | GPT-6 Astra (max) |
Room left to improve
Nearly maxed out
Best score 96% 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
PhD-level experts answering questions in their own field score about 81% on this set. Skilled people from other fields, with unrestricted web access and around half an hour per question, score about 22%.
SourceWhat it tests
Multiple-choice questions in biology, chemistry and physics, written by people with or working towards PhDs and checked so the answer can’t simply be searched for online.
Why it matters
It shows whether a model can reason through graduate-level science rather than recall facts.
Caveats
Top models now sit near the ceiling, so small differences between them mean little. Four answer options also mean a random guess scores about 25%, which is why the line never starts at zero.
How benchmarks work →Model scores
Top 10 of 50 models by score
| Rank | Model | Developer | Score | Released |
|---|---|---|---|---|
| 1 | GPT-6 Astra (max) OpenAI | OpenAI | 96% | Sep 2026 |
| 2 | Claude Sonnet 5.5 (max) Anthropic | Anthropic | 96% | Sep 2026 |
| 3 | Gemini 3.8 Flash (high) Google DeepMind | Google DeepMind | 95% | Sep 2026 |
| 4 | GPT-6.1 Sol (max) OpenAI | OpenAI | 95% | Sep 2026 |
| 5 | Gemini 3.7 Flash (high) Google DeepMind | Google DeepMind | 95% | Aug 2026 |
| 6 | GPT-5.4 Pro (xhigh) OpenAI | OpenAI | 95% | Mar 2026 |
| 7 | Gemini 3.1 Pro Preview Google DeepMind | Google DeepMind | 94% | Feb 2026 |
| 8 | GPT-6 Sol (max) OpenAI | OpenAI | 94% | Sep 2026 |
| 9 | Gemini 3.6 Flash Google DeepMind | Google DeepMind | 94% | Jul 2026 |
| 10 | GPT-5.5 (xhigh) OpenAI | OpenAI | 94% | Apr 2026 |
Benchmark data from Epoch AI (CC BY 4.0) and official leaderboards. Updated weekly, last on . Download CSV Methodology