K-Bench (Therapeutic v0): leaderboard
K-Bench overall clinician-rubric score for high-risk mental-health vignettes using the Therapeutic v0 prompt. Published reasoning settings remain separate from each other and from the Default v1 prompt board. Higher is better. Source: arxiv.org/abs/2609.15855.
Metric: Overall (%, Therapeutic v0 prompt). Source: arxiv.org. Saturation forecast: Estimated already saturated. 33 models tracked. Updated 4 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5.2 | Non-reasoning | 98.95 | #187 |
| 2 | GPT-5.5 | Low | 98.72 | #54 |
| 3 | GPT-5.4 | Low | 98.51 | #138 |
| 4 | Kimi K2.5 | 98.45 | #265 | |
| 5 | Claude Haiku 4.5 | Non-reasoning | 98.45 | #586 |
| 6 | Kimi K2.6 | 98.11 | #192 | |
| 7 | Claude Fable 5 | Low | 98.11 | #32 |
| 8 | GPT-5.6 Terra | Non-reasoning | 98.05 | #72 |
| 9 | GLM-5.3 Flash | Low | 97.96 | #136 |
| 10 | Claude Opus 4.8 | Low | 97.87 | #64 |
| 11 | Nemotron 3 Ultra 550B A55B | High | 97.83 | |
| 12 | Gemini 3.5 Flash | Medium | 97.71 | #108 |
| 13 | Qwen 3 32B | High | 97.65 | #972 |
| 14 | GPT-5.6 Sol | Non-reasoning | 97.62 | #36 |
| 15 | DeepSeek V4 Flash | High | 97.61 | #267 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=k-bench-therapeutic-v0 · How It Works · Data refreshed daily, snapshot 2026-11-20.