OpenAI MentalHealthBench - Acuity - Non-acute: leaderboard
Acuity split of OpenAI's MentalHealthBench (launch post and paper, 2026-09-23): the 650 non-acute conversations (everyday concerns with an emotional component). 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 7.
Metric: Task-clipped rubric score (%). Source: openai.com. Saturation forecast: Around 2030. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 56.64 | #5 |
| 2 | GPT-6 Sol | 52.26 | #17 |
| 3 | Claude Opus 5.5 | 51.42 | #3 |
| 4 | GPT-6 Luna | 47.88 | #62 |
| 5 | GPT-5.6 Sol (August 2026) | 45.21 | |
| 6 | Claude Fable 5.1 | 44.19 | #10 |
| 7 | Claude Sonnet 5 | 43.9 | #67 |
| 8 | Muse Spark 1.3 | 43.21 | #31 |
| 9 | GPT-5.6 Luna (August 2026) | 43.01 | |
| 10 | Claude Haiku 4.5 | 41.4 | #291 |
| 11 | Grok 4.7 | 40.16 | #26 |
| 12 | GPT-5 (Thinking) | 39.7 | #108 (GPT-5) |
| 13 | GPT-4o (Mar 2025) | 36.46 | #339 |
| 14 | Gemini 2.5 Flash | 34.1 | #255 |
| 15 | Gemini 3.8 Flash | 32.28 | #30 |
No result here: #6 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=openai-mentalhealthbench-acuity-non-acute · How It Works · Data refreshed daily, snapshot 2026-11-07.