SEAL - Humanity's Last Exam (Text Only): leaderboard
2,500 expert-vetted text-only questions spanning graduate-level mathematics, humanities, and sciences at the frontier of AI difficulty. Current frontier models score below 10%.
Metric: Score. Source: scale.com. Saturation forecast: Around November 2027. 65 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 54.17 | #6 |
| 2 | Gemini 3.1 Pro (Preview) (High) | 47.31 | #58 (Gemini 3.1 Pro (Preview)) |
| 3 | Claude Fable 5.1 (xHigh) | 46.8 | #9 (Claude Fable 5.1) |
| 4 | Claude Sonnet 5.5 (xHigh) | 46.62 | #11 (Claude Sonnet 5.5) |
| 5 | Gemini 3.8 Flash | 44.95 | #21 |
| 6 | Muse Spark | 40.92 | #64 |
| 7 | Gemini 3 Pro (Preview) | 37.72 | #66 |
| 8 | Claude Opus 4.6 (Adaptive Reasoning, Max Effort) | 36.24 | #63 (Claude Opus 4.6) |
| 9 | Claude Haiku 5.5 | 30.35 | #40 |
| 10 | GPT-5.2 | 28.5 | #110 |
| 11 | GPT-5 | 26.32 | #98 |
| 12 | Claude Opus 4.5 (20251101) (Thinking) | 26.32 | #74 (Claude Opus 4.5 (20251101)) |
| 13 | GPT-5.1 (Thinking) | 24.65 | #135 (GPT-5.1) |
| 14 | Gemini 2.5 Pro (Preview 06-05) | 22.06 | #106 |
| 15 | O3 (High) | 20.57 | #124 (O3) |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=seal-humanity-s-last-exam-text-only · How It Works · Data refreshed daily, snapshot 2026-10-20.