HAL GAIA — leaderboard
Metric: Accuracy (%). Source: hal.cs.princeton.edu. 32 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.5 | 74.55 |
| 2 | Claude Sonnet 4.5 (High) | 70.91 |
| 3 | Claude Opus 4.1 (High) | 68.48 |
| 4 | Claude Opus 4.1 (20250805) | 64.24 |
| 5 | Claude 3.7 Sonnet (High) | 64.24 |
| 6 | GPT-5 (Medium) | 62.8 |
| 7 | O4 Mini (Low) | 58.18 |
| 8 | Claude Opus 4 (20250514) | 57.58 |
| 9 | Claude Haiku 4.5 (20251001) | 56.36 |
| 10 | Claude 3.7 Sonnet (20250219) | 56.36 |
| 11 | O4 Mini (High) | 55.76 |
| 12 | GPT-4.1 (2025-04-14) | 50.3 |
| 13 | Gemini 2.0 Flash (001) | 32.73 |
| 14 | O3 (Medium) | 32.73 |
| 15 | DeepSeek R1 | 30.3 |
Interactive version: theaggregate.ai/benchmark?slug=hal-gaia · How the rankings work · Data refreshed daily, snapshot 2026-07-22.