K-Bench - Risk (Therapeutic v0): leaderboard
K-Bench risk-handling score using the Therapeutic v0 prompt, with each published reasoning setting preserved. Measures management of danger to self or others against the clinician-agreed rubric. Higher is better. Source: arxiv.org/abs/2609.15855.
Metric: Risk Score (%, 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 | Claude Fable 5 | Low | 95.17 | #32 |
| 2 | GPT-5.5 | Non-reasoning | 95.16 | #54 |
| 3 | GPT-5.2 | Non-reasoning | 95.15 | #187 |
| 4 | GPT-5.4 | Low | 94.38 | #138 |
| 5 | GPT-5.6 Terra | Non-reasoning | 94.25 | #72 |
| 6 | GPT-5.6 Luna | Non-reasoning | 93.78 | #133 |
| 7 | GPT-5.6 Sol | Non-reasoning | 93.64 | #36 |
| 8 | Claude Haiku 4.5 | Non-reasoning | 93.41 | #586 |
| 9 | Kimi K2.5 | 92.92 | #265 | |
| 10 | Claude Opus 4.8 | Low | 92.5 | #64 |
| 11 | GLM-5.3 Flash | Low | 91.83 | #136 |
| 12 | Kimi K2.6 | 91.2 | #192 | |
| 13 | Nemotron 3 Ultra 550B A55B | High | 90.86 | |
| 14 | GLM-5.2 | High | 90.72 | #155 |
| 15 | Gemini 3.5 Flash | Medium | 90.43 | #108 |
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-risk-therapeutic-v0 · How It Works · Data refreshed daily, snapshot 2026-11-20.