GPT-6 Cyber

OpenAI launched GPT-6 Cyber, a restricted cybersecurity model rated Critical under its Preparedness Framework.

Announced Nov 5 for approved Daybreak Red customers via the API and Codex. OpenAI says it scores 48.5% on ExploitGym (up from 32.1% for GPT-5.6 Cyber), stays in scope in nearly all honeypot tests, and helped find 81 high-severity Firefox bugs and exploit 948 Linux privilege-escalation issues in research campaigns. Access starts with a select Daybreak Red group and expands later; GPT-5.6 Cyber is opening to all Daybreak Red customers.

Status

Released

Date

Nov 5, 2026 (Daybreak Red, limited)

Maker

OpenAI

Type

AI model

Price

Daybreak Red (vetted customers); expanding soon

Categories

AI

Sources

• OpenAI
• The Verge
• Reuters via Investing.com

Launch dates can change. Check the sources for the latest. Last reviewed Nov 10, 2026.