OpenAI MentalHealthBench - Behavior - User Agency: leaderboard
Behavior axis of OpenAI's MentalHealthBench (launch post and paper, 2026-09-23): respecting autonomy and supporting the user's own informed decisions, scored on the 170 rubric criteria tagged with this axis. The post's axis score is the signed rubric score on this axis, normalized to its possible range (0-100, higher is better); GPT-5.6 Sol grades, models run at API default reasoning effort. The post's 15 models only: the paper's behavior figure uses a different scale.
Metric: Axis score, normalized to its possible range (%). Source: openai.com. Saturation forecast: Not forecast. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 5 | 67.09 | #67 |
| 2 | Claude Opus 5.5 | 63.54 | #3 |
| 3 | Muse Spark 1.3 | 62.69 | #31 |
| 4 | Claude Haiku 4.5 | 61.4 | #291 |
| 5 | GPT-4o (Mar 2025) | 59.51 | #339 |
| 6 | GPT-6 Astra | 59.18 | #5 |
| 7 | Claude Fable 5.1 | 58.49 | #10 |
| 8 | GPT-6 Sol | 58.3 | #17 |
| 9 | GPT-6 Luna | 58.26 | #62 |
| 10 | Grok 4.7 | 55.82 | #26 |
| 11 | GPT-5.6 Sol (August 2026) | 54.51 | |
| 12 | GPT-5.6 Luna (August 2026) | 53.02 | |
| 13 | Gemini 3.8 Flash | 50.43 | #30 |
| 14 | Gemini 2.5 Pro | 49.03 | #161 |
| 15 | Gemini 3.1 Pro (Preview) | 45.4 | #66 |
No result here: #6 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=openai-mentalhealthbench-behavior-user-agency · How It Works · Data refreshed daily, snapshot 2026-11-07.