MentalHealthBench
Safety · 2026-09-23
OpenAI's benchmark of responses in 1,215 synthetic mental-health conversations across non-acute, high-acuity and emergent situations, adult, teen, clinician and caregiver users and 11 languages, graded against 5,262 rubric criteria written by more than 80 licensed psychiatrists and psychologists. A model grader marks each criterion met or not; each response's signed rubric score is clipped at zero and averaged over four sampled responses and all conversations. Higher is better.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 57.3 |
| GPT-6 Sol | 53.9 |
| Claude Opus 5.5 | 52.4 |
| GPT-6 Luna | 50.2 |
| Muse Spark 1.3 | 48.6 |
| GPT-5.6 Sol (ChatGPT, August 2026) | 47.0 |
| Claude Fable 5.1 | 46.4 |
| GPT-5.6 Luna (ChatGPT, August 2026) | 44.9 |
| Sonnet 5 | 44.5 |
| GPT-5 | 42.9 |
| Haiku 4.5 | 41.7 |
| Grok 4.7 | 41.3 |
| Gemini 3.8 Flash | 35.5 |
| Gemini 2.5 Flash | 33.5 |
| ChatGPT-4o Latest observed 2025-03-27 | 32.1 |
| Gemini 3.1 Pro Preview | 32.1 |
| Gemini 2.5 Pro | 29.5 |