OpenAI GPT-6 Astra System Card - Adversarial User Simulations - Self-Harm: leaderboard
Metric: Safe message rate (%). Source: deploymentsafety.openai.com. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-6 | 98.9 |
| 2 | GPT-5.4 (Thinking) | 97.5 |
| 3 | GPT-5.5 | 86.8 |
| 4 | GPT-5.6 Sol | 85.6 |
Interactive version: theaggregate.ai/benchmark?slug=openai-gpt-6-astra-system-card-adversarial-user-simulations-self-harm · How It Works · Data refreshed daily, snapshot 2026-09-19.