Agent Security League - Functional Correctness: leaderboard
Metric: Functional Correctness (%). Source: www.endorlabs.com. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 84.9 |
| 2 | Claude Opus 4.6 | 84.9 |
| 3 | Claude Sonnet 5 | 83.2 |
| 4 | GLM-5.2 | 82.5 |
| 5 | Claude Opus 4.7 | 79.9 |
| 6 | Gemini 3.5 Flash | 79.3 |
| 7 | Claude Opus 4.8 | 75.4 |
| 8 | Composer 2.5 | 75.4 |
| 9 | Claude Opus 5 | 73.7 |
| 10 | Gemini 3.1 Pro (Preview) | 73.2 |
| 11 | Claude Fable 5 | 72.6 |
| 12 | GPT-5.6 Sol | 70.9 |
| 13 | Claude Sonnet 4.6 | 69.8 |
| 14 | Claude Opus 4.5 | 68.7 |
| 15 | GPT-5.4 | 63.1 |
Interactive version: theaggregate.ai/benchmark?slug=agent-security-league-functional-correctness · How It Works · Data refreshed daily, snapshot 2026-09-05.