Playbooks · Teams tasks
Daybreak Red reaches AWS GovCloud
Written November 10, 2026
GPT-5.6 Cyber, the model behind Daybreak Red, is now available to approved customers in AWS GovCloud (US-East) through Amazon Bedrock. Approval comes first; the region comes second.
What it is
AWS announced on October 30 that Daybreak Red from OpenAI is available to eligible customers on Amazon Bedrock in AWS GovCloud (US-East); OpenAI's own announcement followed on November 4. The model is GPT-5.6 Cyber (Bedrock model ID openai.gpt-5.6-cyber), built for authorized vulnerability research, exploit reproduction, and mitigation development. It is available in US East (Ohio) and AWS GovCloud (US-East), with In-Region inference in us-gov-east-1 and us-east-2; Geo and Global cross-Region inference are not supported for this model.
Daybreak Red runs on Bedrock's next-generation inference engine, with zero-operator access enforced at the chip. Inference data is not used to train models, and customers do not need to share their Daybreak Red data with OpenAI. The context window is 1,050,000 tokens.
Two gates, in order
Access takes two approvals, and the second does not follow automatically from the first. An organisation first needs approval through OpenAI's Trusted Access for Cyber program — the Daybreak Red enrollment, with identity verification and use restrictions. Then it must ask its AWS account team to enable access in the AWS account. OpenAI approval alone does not activate the model in Amazon Bedrock. Simply having an AWS GovCloud account does not give access.
What it costs
AWS lists Standard-tier In-Region rates for short-context requests of 272,000 input tokens or fewer. In AWS GovCloud (US-East): $16.875 per million input tokens, $1.6875 per million cache-read tokens, and $101.25 per million output tokens. The same band in commercial AWS In-Region is $13.75 / $1.375 / $82.50 — GovCloud runs about 23% higher. The 1,050,000-token context window is the model's limit; the listed prices apply only to requests in the short-context band.
What it is not
This announcement covers GPT-5.6 Cyber only — not GPT-6 Cyber, the newer model, which travels a different path. Daybreak Red is not a public API and not open to general users: it is a gated program for approved defenders doing authorized work, and the announcements name no government customer deployment. GPT-5.6 Cyber was already available through Bedrock in US East (Ohio) since August; the GovCloud move changes the operating environment and the compliance boundary, not the model.
Limits
- Approved customers, authorized work only. Eligibility criteria and pricing beyond the published Standard-tier rates are not set out in the announcements.
- OpenAI says qualified security practitioners should review findings and proposed fixes before implementation — the model assists the work, it does not sign it off.
- Neither announcement names a customer deployment or gives a capability evaluation. The news is availability in the region, not a measured result.
Related playbooks
- ChatGPT Enterprise joins AWS Marketplace
- Iterable's hosted MCP server reaches your AI tools
- Budget agentic autofix before you enable it
- Anthropic's model-protection rule is now in force
Sources
- Daybreak Red from OpenAI now available to eligible customers on Amazon Bedrock in AWS GovCloud (US-East) — AWS What's New
- Daybreak Red brings GPT-5.6 Cyber to AWS GovCloud — NeoTeo
Back to: Small-team playbooks hub · Also: ChatGPT Enterprise joins AWS Marketplace · GPT-6 Cyber: OpenAI's restricted security model