GPT-6 Cyber is an OpenAI model for security work: it helps approved teams look into software flaws, recreate them safely and check that fixes hold.
In plain words
It was added to OpenAI's API on 5 November 2026. Security teams use it to investigate a weakness, rebuild it inside a test setup they are allowed to attack, and confirm a patch works. Access is restricted: an organisation needs approval for Daybreak Red, OpenAI's vetted programme for defenders, plus a separate approval for this model. It reads text and images, answers in text and can take in about a million tokens at once.
A quick example
Imagine a hospital's security team hears about a flaw in software it runs. Working in a copy of its own system set aside for testing, the team asks the model to confirm the flaw is real, then to check whether the vendor's patch closes it.
Where you'll meet it
- Vulnerability research by approved security teams.
- Checking that security fixes actually work.
Related
Sources
- DocsGPT-6-Cyber model, OpenAI · read 12 Nov 2026
- DocsAPI changelog, OpenAI · read 12 Nov 2026
- DocsDaybreak guide, OpenAI · read 12 Nov 2026
Checked by WhatIsAI against the sources above · 12 Nov 2026. Spotted an error? Report it.