OpenAI MentalHealthBench - User Profile - Caregiver: leaderboard
User-profile split of OpenAI's MentalHealthBench (launch post and paper, 2026-09-23): the 60 conversations with a caregiver user. Mean task-clipped rubric score, 0-100, higher is better: GPT-5.6 Sol grades each reply against expert-written signed rubrics, and models run at their API default reasoning effort. GPT-5 Thinking and Gemini 2.5 Flash come from the paper's Figure 10.
Metric: Task-clipped rubric score (%). Source: openai.com. Saturation forecast: Around March 2028. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 66.3 | #6 |
| 2 | GPT-6 Sol | 64.6 | #24 |
| 3 | Claude Opus 5.5 | 62.98 | #5 |
| 4 | Muse Spark 1.3 | 61.51 | #42 |
| 5 | GPT-5.6 Sol (August 2026) | 60.31 | |
| 6 | Claude Fable 5.1 | 57.6 | #14 |
| 7 | GPT-5 (Thinking) | 57 | #163 (GPT-5) |
| 8 | GPT-6 Luna | 56.9 | #93 |
| 9 | GPT-5.6 Luna (August 2026) | 55.77 | |
| 10 | Claude Sonnet 5 | 55.7 | #108 |
| 11 | Grok 4.7 | 52.24 | #43 |
| 12 | Claude Haiku 4.5 | 51.97 | #568 |
| 13 | Gemini 3.8 Flash | 45.09 | #36 |
| 14 | Gemini 2.5 Flash | 44.3 | #452 |
| 15 | Gemini 3.1 Pro (Preview) | 41.6 | #91 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #10 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openai-mentalhealthbench-user-profile-caregiver · How It Works · Data refreshed daily, snapshot 2026-11-12.