K-Bench - Substance Misuse (Therapeutic v0): leaderboard
The model talks with a simulated patient presenting substance misuse for 20 turns under the Therapeutic v0 system prompt; the number is the 0-100 risk-handling score (clinical judgement and risk exploration) on that domain's vignettes. Unlike the K-Bench - Risk board, which pools the four risk domains, it shows the domain alone.
Metric: Risk-handling score in this domain (0-100, higher is better). Source: k-bench.ai. 61 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (Non-reasoning) | 95.45 | #176 (GPT-5.2) |
| 2 | Claude Fable 5 (Low) | 95.41 | #27 (Claude Fable 5) |
| 3 | GPT-5.5 (Non-reasoning) | 95.05 | #49 (GPT-5.5) |
| 4 | GPT-5.6 Terra (Non-reasoning) | 94.97 | #62 (GPT-5.6 Terra) |
| 5 | GPT-5.2 (High) | 94.74 | #176 (GPT-5.2) |
| 6 | GPT-5.6 Terra (High) | 94.71 | #62 (GPT-5.6 Terra) |
| 7 | GPT-5.4 (Low) | 94.55 | #123 (GPT-5.4) |
| 8 | GPT-5.6 Terra (Medium) | 94.46 | #62 (GPT-5.6 Terra) |
| 9 | GPT-5.5 (High) | 94.32 | #49 (GPT-5.5) |
| 10 | GPT-5.4 (High) | 94.02 | #123 (GPT-5.4) |
| 11 | GPT-5.6 Sol (Non-reasoning) | 93.92 | #30 (GPT-5.6 Sol) |
| 12 | GPT-5.6 Luna (Non-reasoning) | 93.85 | #121 (GPT-5.6 Luna) |
| 13 | GPT-5.4 (Non-reasoning) | 93.58 | #123 (GPT-5.4) |
| 14 | GPT-5.5 (Medium) | 93.56 | #49 (GPT-5.5) |
| 15 | Claude Fable 5 (High) | 93.4 | #27 (Claude Fable 5) |
Interactive version: theaggregate.ai/benchmark?slug=k-bench-substance-misuse-therapeutic-v0 · How It Works · Data refreshed daily, snapshot 2026-11-13.