HumanAgencyBench - Correct Misinformation: leaderboard
Metric: HAB dimension score (0-100, o3-judged rubric). Source: github.com. 38 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 88.56 |
| 2 | Claude Opus 4.6 | 86.04 |
| 3 | GPT-5 (High) | 85.28 |
| 4 | GPT-5.5 | 84.1 |
| 5 | Claude Opus 4.5 (20251101) | 82.32 |
| 6 | Claude Sonnet 4.6 | 79.72 |
| 7 | GPT-5.4 | 79.66 |
| 8 | Claude Opus 4.1 (20250805) | 79.18 |
| 9 | GPT-5 | 73.28 |
| 10 | Claude Sonnet 4 (20250514) | 70.64 |
| 11 | Gemini 3.1 Pro (Preview) | 67.18 |
| 12 | Grok 4 | 66.84 |
| 13 | Kimi K2.6 | 65.48 |
| 14 | GPT-5.4 Mini | 60.88 |
| 15 | O3 (2025-04-16) | 59.48 |
Interactive version: theaggregate.ai/benchmark?slug=humanagencybench-correct-misinformation · How It Works · Data refreshed daily, snapshot 2026-09-19.