HumanAgencyBench - Encourage Learning: leaderboard
HumanAgencyBench dimension Encourage Learning: 500 prompts measuring how well the assistant promotes the user's learning and intellectual growth rather than only supplying answers. 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) | 78.84 | #81 |
| 2 | Claude Sonnet 4.6 | 73.74 | #94 |
| 3 | Claude Opus 4.7 | 70.98 | #51 |
| 4 | Claude Opus 4.6 | 69.22 | #67 |
| 5 | Claude Haiku 4.5 (20251001) | 68 | #264 |
| 6 | Claude Opus 4.1 (20250805) | 61.98 | #139 |
| 7 | Claude Sonnet 4 (20250514) | 57.52 | #223 |
| 8 | Gemini 3.1 Pro (Preview) | 54.5 | #64 |
| 9 | Grok 3 | 48.3 | #313 |
| 10 | Claude 3.5 Sonnet (20241022) | 48.26 | #302 |
| 11 | Gemini 2.5 Pro (Preview 03-25) | 47.86 | #126 |
| 12 | DeepSeek V4 Pro | 45.28 | #112 |
| 13 | GPT-5.4 | 43.92 | #83 |
| 14 | Kimi K2.6 | 43.4 | #113 |
| 15 | GPT-5 | 42.92 | #105 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=humanagencybench-encourage-learning · How It Works · Data refreshed daily, snapshot 2026-11-05.