GPT-6 Cyber launches for selected Daybreak Red customers
OpenAI announced GPT-6 Cyber on November 5, initially through the API and Codex for selected approved Daybreak Red customers. Security teams should confirm that their specific account has been provisioned for the new model.
Access will expand to all Red customers soon, without a stated date. OpenAI is also widening GPT-5.6 Cyber availability to all Red customers and GPT-6 Astra’s Daybreak Blue alpha audience.
Who can use GPT-6 Cyber now?
The named launch audience excludes automatic access for every ChatGPT subscriber or Daybreak member. The announcement does not establish a general ChatGPT rollout or access through other cloud providers.
Daybreak is OpenAI’s access program for authorized cybersecurity work. Its application page accepts individuals and organizations, with applicant verification and review of intended use. Applying is the start of a review, not a promise of immediate model access.
The new model follows GPT-5.6 Cyber’s earlier Daybreak Red rollout. That background explains the program, but an older Red deployment should not be treated as proof that GPT-6 Cyber is enabled in the same account.
Approval and activation are separate checks
OpenAI’s enterprise onboarding guide distinguishes eligibility, provisioning and configuration. GPT-6 Cyber requires separate model approval in addition to Daybreak Red access. Teams should check the models, users and product surfaces named in their onboarding confirmation.
The guide also distinguishes workspace access from API project access. Enabling one does not configure the other. An API organization owner manages approved non-default projects, while a workspace owner manages approved users and groups. A missing model or control should be checked with the account team before moving work between projects.
For a team planning adoption, the useful checklist is short: confirm GPT-6 Cyber is included in the approval, confirm the intended workspace or project, and confirm that the people running it have the corresponding access. Keep those checks separate from evaluating whether the model suits the work.
The onboarding guide also lists enforced single sign-on for Daybreak Red organizations and workspaces, default IP allowlisting for GPT-6 Cyber API access, and Private Safety Processing generally required for API projects using Zero Data Retention. These are access requirements to review with the administrator.
What restrictions still apply?
The Daybreak access overview limits use to approved internal users and workflows. It does not authorize testing systems without permission, reselling access or passing third-party customer traffic through the program. Approval does not remove every safeguard or guarantee every cyber-specialized model.
Data retention also needs its own check. The overview says Zero Data Retention is separate from Trusted Access. Teams with contractual retention requirements should verify their approved arrangement rather than infer it from the Daybreak tier.
Codex pricing is published, with a limited scope
OpenAI’s updated Enterprise token rate card lists GPT-6 Cyber in Codex at $20 per million input tokens, $2 per million cached input tokens and $100 per million output tokens. Those Standard-mode rates apply to eligible token-based Enterprise agreements and approved GPT-6 Cyber access. Fast and Ultrafast modes are not supported.
The Codex row does not establish pricing for every API customer. Check the agreement and billing path attached to the provisioned account before estimating a project budget. Broader Red access still has no firm rollout date in the launch announcement.
