GPT-6 Cyber Guide: Daybreak Red Access and Pricing

On November 5, 2026, OpenAI released GPT-6 Cyber, a cyber-specialized model for authorized vulnerability research under Daybreak Red. The API model ID is gpt-6-cyber. Access needs Daybreak Red plus separate model approval, with $20 input and $100 output per million tokens on the published short-context rates.

Last updated: November 6, 2026

What shipped on November 5, 2026
Daybreak Red access and who can apply
Model specs, pricing and endpoints
How to call gpt-6-cyber
Safeguards and operational limits
Frequently asked questions
Sources

What shipped on November 5, 2026

OpenAI’s GPT-6 Cyber announcement and the API changelog introduce gpt-6-cyber for investigating vulnerabilities, reproducing issues in an authorized test environment, and verifying fixes. OpenAI positions the model as a step up from GPT-5.6 Cyber on internal cyber evaluations such as ExploitGym, with stronger emphasis on staying inside the researcher’s stated scope.

At a high level, OpenAI reports that GPT-6 Cyber is classified as Critical in cybersecurity under its Preparedness Framework (same designation as GPT-6 Astra), and that it is initially available through the API and Codex to a select group of approved Daybreak Red customers, with broader Daybreak Red rollout planned. Mainline models such as GPT-6 Astra remain available for code review and triage; Daybreak Blue and Daybreak Red control how cyber-oriented requests are treated.

For related OpenAI product coverage on FreeAnswers, see the Codex CLI 0.167 power-user guide, What is GPT-6 Astra?, and OpenAI API pricing.

Daybreak Red access and who can apply

OpenAI’s Daybreak Help Center overview describes Daybreak as Trusted Access for Cyber. Daybreak Blue is the usual starting point for defensive workflows on supported mainline models. Daybreak Red is for advanced, authorized workflows and is currently available only to approved business and enterprise organizations—not through the individual Blue application path.

Important access facts from OpenAI’s docs:

All Daybreak Red organizations are eligible for GPT-5.6 Cyber.
GPT-6 Cyber requires Daybreak Red plus separate model approval.
Turning on Daybreak Red in a workspace or API project does not by itself grant GPT-6 Cyber.
Daybreak must not be enabled for customer-facing applications, downstream product traffic, or third-party access.

Organizations and individuals apply through OpenAI’s Daybreak request flows. Approval is not automatic and depends on verification, intended use, and trust review. Individuals can apply for Daybreak Blue (with Advanced Account Security and a physical FIDO2 hardware key among other requirements); Daybreak Red remains an organization path.

Model specs, pricing and endpoints

Figures below come from OpenAI’s GPT-6-Cyber model page (checked this run).

Item	GPT-6 Cyber (published)
Model ID	`gpt-6-cyber`
Access	Daybreak Red + additional model approval
Context window	1,050,000 tokens
Max input / max output	922,000 / 128,000 tokens
Modalities	Text and image in; text out
Input / cached input / output	$20 / $2 / $100 per 1M tokens
Cache writes	$25 per 1M (30 min) or $50 per 1M (12 h)
Primary endpoints	Responses and Chat Completions supported; Batch not supported

Those are standard short-context prices. OpenAI’s Daybreak guide maps GPT-6-Cyber to the request value daybreak_red. For plan-level consumer pricing context outside Daybreak, see How much does ChatGPT Plus cost? and ChatGPT Images 2.6.

How to call gpt-6-cyber

Follow OpenAI’s Use Daybreak in the Responses API guide. The request parameter selects behavior within approved access; it does not grant access by itself.

Confirm with your account team that the organization has Daybreak Red and GPT-6 Cyber approval.
Require SSO on the Cyber API organization and enforce MFA through your identity provider.
Ask an organization owner to enable Daybreak Red on a non-default project; create an API key in that project.
Add your public egress IP addresses to the project or organization IP allowlist (required by default for Daybreak Red API access to GPT-6 Cyber).
If the project uses Zero Data Retention, configure Private Safety Processing and verify the retention policy shows “Zero Data Retention with Private Safety Processing.”
Call the Responses API with model set to gpt-6-cyber, access_programs.cyber set to daybreak_red, and a stable pseudonymous safety_identifier. Use only authorized, in-scope security tasks.

Common API errors from the same guide: 400 invalid_access_program (wrong program value for the model), 400 unsupported_access_program (model does not support Daybreak), and 403 access_program_not_enabled (project or org missing the required program). In Codex signed in with ChatGPT, use the approved workspace and enable Daybreak where the client provides a control; API project enablement and workspace access are separate.

Safeguards and operational limits

OpenAI’s GPT-6 Cyber system card / Deployment Safety Hub and announcement describe layered controls: identity verification, use restrictions, SSO/MFA, API IP allowlisting, Private Safety Processing for many ZDR setups, cyber-specific misalignment monitoring, and checks aimed at high-risk assistance involving critical infrastructure. Daybreak does not remove all refusals, does not guarantee every cyber model, and does not include Zero Data Retention by default.

Use GPT-6 Cyber only on systems you own or are explicitly authorized to test. OpenAI’s Cyber Abuse Policy disallows compromise of systems without proper authorization or beyond granted scope. Prefer sandboxed environments, action review, and human oversight for high-risk workflows—OpenAI recommends comparable controls for API agents when Codex sandboxing and Auto-Review are not in the path.

This guide covers product access, API setup, model identifiers, published limits, and pricing. It does not provide exploit development steps, proof-of-concept attack procedures, or vulnerability reproduction walkthroughs.

Frequently asked questions

What is GPT-6 Cyber?

GPT-6 Cyber is OpenAI’s cyber-specialized variant of GPT-6 Astra for authorized security work. OpenAI’s November 5, 2026 announcement and system card describe stronger results than GPT-5.6 Cyber on internal cyber benchmarks, Critical cybersecurity preparedness designation, and availability through Daybreak Red with separate model approval.

How do I get access to GPT-6 Cyber?

Your organization needs Daybreak Red approval and an additional approval specifically for GPT-6 Cyber. Daybreak Red is for approved business and enterprise organizations, not the individual Daybreak Blue application path. After approval, an organization owner enables Daybreak Red on a non-default API project, and you use a key from that project.

What is the API model ID and request parameter?

Set model to gpt-6-cyber and access_programs.cyber to daybreak_red on the Responses API. Selecting Daybreak Red alone does not switch your model or grant GPT-6 Cyber. OpenAI’s Daybreak guide also recommends a stable safety_identifier for the end user making the request.

How much does GPT-6 Cyber cost?

OpenAI’s model page lists standard short-context prices of $20 per 1M input tokens, $2 per 1M cached input tokens, $25 per 1M cache writes (30-minute retention; $50 for 12-hour), and $100 per 1M output tokens. Confirm long-context or custom rates with your account team if needed.

What setup is required before the first API call?

Daybreak Red organizations must require SSO and MFA on the Cyber API organization. For GPT-6 Cyber specifically, configure an IP allowlist for your egress IPs. If the project uses Zero Data Retention, set up Private Safety Processing and confirm storage shows Validated with that retention policy.

How does GPT-6 Cyber differ from Daybreak Blue and GPT-5.6 Cyber?

Daybreak Blue reduces refusals on supported mainline models for defensive work and does not include cyber-specialized models. Daybreak Red adds models such as GPT-5.6 Cyber for all Red organizations; GPT-6 Cyber needs Red plus separate model approval. GPT-6 Cyber is the newer cyber-specialized model with a larger published context window and higher listed token prices than GPT-5.6 Cyber.

GPT-6 Cyber: Power-User Guide to Daybreak Red Access and API

Table of contents