OpenAI MentalHealthBench - Acuity - Emergent: leaderboard
Acuity split of OpenAI's MentalHealthBench (launch post and paper, 2026-09-23): the 344 emergent conversations (signs of a mental health emergency or immediate safety risk). 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 2028. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 58.26 | #5 |
| 2 | GPT-6 Sol | 55.14 | #22 |
| 3 | Muse Spark 1.3 | 54.35 | #45 |
| 4 | Claude Opus 5.5 | 53.91 | #4 |
| 5 | GPT-6 Luna | 52.61 | #90 |
| 6 | Claude Fable 5.1 | 47.77 | #13 |
| 7 | GPT-5.6 Sol (August 2026) | 46.08 | |
| 8 | GPT-5 (Thinking) | 44.2 | #164 (GPT-5) |
| 9 | GPT-5.6 Luna (August 2026) | 43.67 | |
| 10 | Claude Sonnet 5 | 43.1 | #108 |
| 11 | Grok 4.7 | 41.55 | #41 |
| 12 | Claude Haiku 4.5 | 41.09 | #593 |
| 13 | Gemini 3.8 Flash | 37.86 | #37 |
| 14 | Gemini 3.1 Pro (Preview) | 31.78 | #93 |
| 15 | Gemini 2.5 Flash | 29.6 | #464 |
No result here: #9 GPT-6.1 Sol, #10 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=openai-mentalhealthbench-acuity-emergent · How It Works · Data refreshed daily, snapshot 2026-11-08.