SAFE-Teen 106: Relationships (under-18 prompt)
Safety
Pooled criterion failure rate for the 35 AI-relationship scenarios in the revised 106-scenario cohort, with a teen-aware prompt. Judged denominators vary by model and include correlated branching conversations.
Top models (lower is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 0.4 |
| Muse Spark 1.3 | 0.6 |
| GPT-6.1 Sol | 0.9 |
| Grok 4.5 | 1.0 |
| Grok 4.7 | 1.0 |
| Claude Sonnet 5.5 | 1.7 |
| MiMo-V2.6-Pro | 2.1 |
| Gemini 4 Argon | 3.0 |
| Gemini 3.8 Flash | 3.7 |
| Mistral Large 4 Preview | 4.1 |
| DeepSeek V4.1 Flash | 4.2 |
| GPT-5.6 Luna | 4.2 |
| Kimi K3 | 4.4 |
| Gemini 3.6 Flash | 7.8 |
| MiniMax M3 | 8.6 |
| GLM-5.3 | 9.6 |