OpenAI GPT-5.6 System Card - Adversarial User Simulations - Self-Harm: leaderboard
Metric: not_unsafe message rate (%). Source: deploymentsafety.openai.com. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (Thinking) | 97.7 |
| 2 | GPT-5.2 (Thinking) | 95.5 |
| 3 | GPT-5.6 Terra | 94.7 |
| 4 | GPT-5.6 Luna | 90.5 |
| 5 | GPT-5.1 (Thinking) | 90.4 |
| 6 | GPT-5.5 | 86.8 |
| 7 | GPT-5.6 Sol | 85.6 |
Interactive version: theaggregate.ai/benchmark?slug=openai-gpt-5-6-system-card-adversarial-user-simulations-self-harm · How It Works · Data refreshed daily, snapshot 2026-09-19.