HumanAgencyBench - Ask Clarifying Questions: leaderboard
HumanAgencyBench dimension Ask Clarifying Questions: 500 prompts that are ambiguous or missing critical information, testing whether the assistant seeks the missing details instead of proceeding. o3-judged 0-10 rubric score averaged per model and shown on a 0-100 scale; higher is better.
Metric: HAB dimension score (0-100, o3-judged rubric). Source: github.com. Saturation forecast: Around February 2027. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 (20251101) | 90.88 | #81 |
| 2 | Claude Sonnet 4.6 | 83.8 | #94 |
| 3 | Claude Haiku 4.5 (20251001) | 83.48 | #264 |
| 4 | GPT-5 (High) | 81.4 | #105 (GPT-5) |
| 5 | Grok 4 | 78.32 | #189 |
| 6 | GPT-5 | 77.88 | #105 |
| 7 | Claude Opus 4.6 | 70.88 | #67 |
| 8 | Claude 3.5 Sonnet (20241022) | 66.92 | #302 |
| 9 | Claude Opus 4.7 | 62.76 | #51 |
| 10 | Claude Opus 4.1 (20250805) | 62.2 | #139 |
| 11 | Claude Sonnet 4 (20250514) | 49.92 | #223 |
| 12 | GPT-4.1 | 35.32 | #258 |
| 13 | Grok 3 | 33.48 | #313 |
| 14 | Claude 3.7 Sonnet (20250219) | 31.4 | #209 |
| 15 | Kimi K2.6 | 28.08 | #113 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=humanagencybench-ask-clarifying-questions · How It Works · Data refreshed daily, snapshot 2026-11-05.