Agents' Last Exam: leaderboard
Long-horizon agent benchmark covering economically valuable professional workflows in full operating-system sandboxes, with verifiable task outcomes.
Metric: Pass Rate (%). Source: snorkel.ai. Status: saturation imminent. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 | 31.6 |
| 2 | GPT-5.6 Sol | 30.6 |
| 3 | GPT-5.6 Luna | 30.3 |
| 4 | Kimi K3 | 28.3 |
| 5 | GPT-5.6 Terra | 28 |
| 6 | Claude Opus 4.8 | 27 |
| 7 | Grok 4.5 | 27 |
| 8 | GPT-5.5 | 26.6 |
| 9 | Claude Fable 5 | 25.7 |
| 10 | Claude Opus 4.7 | 21.1 |
| 11 | GPT-5.4 | 20.5 |
| 12 | GLM-5.2 | 20.4 |
| 13 | Composer 2.5 | 20.4 |
| 14 | Seed 2.1 Pro | 19.5 |
| 15 | Gemini 3.1 Pro (Preview) | 16.4 |
Interactive version: theaggregate.ai/benchmark?slug=agents-last-exam · How It Works · Data refreshed daily, snapshot 2026-09-05.