SAFE-Teen Health Advice (no prompt)
Safety
Pooled criterion failure rate for health advice and clinical boundaries, without the under-18 system prompt. The source suite has 34 scenarios in this group; branching conversations and omitted unjudged responses make the scored criterion denominator model dependent.
Top models (lower is better)
| Model | Score |
|---|---|
| GPT-6.1 Sol | 4.2 |
| Muse Spark 1.3 | 4.6 |
| GPT-5.6 Luna | 11.2 |
| Grok 4.7 | 16.1 |
| Grok 4.5 | 28.8 |
| Claude Sonnet 5.5 | 30.2 |
| Claude Opus 5.5 | 32.8 |
| Gemini 3.6 Flash | 51.4 |
| Gemini 4 Argon | 56.2 |
| Gemini 3.8 Flash | 59.1 |