GPT-6 Cyber
OpenAI's security-research model, rated Critical for cyber risk, available through API and Codex to select Daybreak Red customers.
Overview
OpenAI unveiled GPT-6 Cyber on November 5, 2026 as a cybersecurity research model that is more capable at finding and fixing vulnerabilities and more reliable at staying inside researchers' defined scope. OpenAI classifies it as Critical in cybersecurity under its Preparedness Framework, and access is limited to select Daybreak Red customers through the API and Codex. OpenAI reports it stayed out of scope in 99.93% of cases without deployment safety prompts (GPT-5.6 Cyber: 94.49%) and scored 48.5% on ExploitGym versus 32.1% for GPT-5.6 Cyber.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Rated Critical in cybersecurity under OpenAI's Preparedness Framework
- 99.93% scope adherence without safety prompts; 100% with them (OpenAI-reported)
- 48.5% on ExploitGym, turning known vulnerabilities into working exploits (GPT-5.6 Cyber: 32.1%)
- OpenAI says it enabled 948 Linux local privilege escalation exploits and 81 new high-severity Firefox flaws
- Includes a critical Brotli vulnerability, CVE-2026-111497 (CVSS 9.2)
- Available via API and Codex to approved Daybreak Red customers only
Use cases
- Authorized vulnerability research and exploit validation
- Reproducing and verifying fixes for known vulnerabilities
- Defender-side red-team and attack simulation work
Pricing
Unknown; not stated in launch coverage. Access is granted to approved Daybreak Red customers rather than sold publicly.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 5 Nov 2026
- Record updated
- 7 Nov 2026
- Record last checked
- 7 Nov 2026