HealthBench-Psych: leaderboard
Rubric-graded answers on 610 mental-health conversations selected from OpenAI's HealthBench with clinical expert review (arXiv 2608.25071v1). Mean of three judges' clipped rubric scores (GPT-4.1, Claude Haiku 4.5 and Gemini 2.5 Flash) as a percentage, higher is better, from eval/runs/matrix_healthbench-psych-v1.csv; unpinned -latest aliases are omitted.
Metric: Three-Judge Panel Mean Rubric Score (%). Source: github.com. Saturation forecast: Around September 2028. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.6 | 62.7 | #113 |
| 2 | GPT-5.5 | 62.39 | #33 |
| 3 | Claude Fable 5.1 | 62.08 | #10 |
| 4 | Claude Opus 5 | 61.96 | #16 |
| 5 | Grok 4.5 | 61.22 | #54 |
| 6 | GPT-5.6 Sol | 60.96 | #22 |
| 7 | Claude Fable 5 | 59.09 | #20 |
| 8 | GPT-6 Astra | 58.48 | #6 |
| 9 | Gemini 3.6 Flash | 57.83 | #57 |
| 10 | Gemini 3.8 Flash | 56.82 | #30 |
| 11 | Kimi K3 | 56.81 | #35 |
| 12 | DeepSeek V4 Pro | 55.43 | #112 |
| 13 | Qwen 3.7 Plus | 55.18 | #98 |
| 14 | DeepSeek V4 Flash | 53.81 | #155 |
| 15 | Claude Sonnet 5 | 53.32 | #65 |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=healthbench-psych · How It Works · Data refreshed daily, snapshot 2026-11-05.