AI Frontier
← Browse this publisher

OpenAI / System Card

GPT-6 Astra System Card

GPT-6 Astra · 2026-09-03

Source summary

Original wording · Original language

1 Safety Overview: GPT-6 Astra · Page 7

Today, we are releasing GPT-6 Astra, the most capable model we have ever broadly deployed. Astra is our first model to reach the Critical level of cybersecurity capability under our Preparedness Framework.

The most important things to know about the safety of this launch are as follows:

Core figures

Enlarge to explore. Download the original for full detail.

Figure 1 · Safety evaluationPage 12
Figure 1
Figure 5 · Safety evaluationPage 20
Figure 5: Indirect prompt injection robustness on the Gray Swan IPI benchmark. Estimated attack

Click the image to zoom. Press Esc to close. Full-resolution files are available below each figure.