GPT-6-Cyber
OpenAI's cybersecurity model for organizations approved for Daybreak Red, built to investigate vulnerabilities, reproduce them in authorized test environments and verify fixes.
OpenAI's cybersecurity model for organizations approved for Daybreak Red, built to investigate vulnerabilities, reproduce them in authorized test environments and verify fixes.
Overview
GPT-6-Cyber is a separate model in OpenAI's Daybreak program, not an access setting on a general model. OpenAI positions it for investigating vulnerabilities, reproducing them in an authorized test environment and verifying fixes. It reads text and images, returns text, and can use tools such as web and file search, code execution, computer use and MCP servers.
Access needs Daybreak Red approval plus a separate approval for this model; Daybreak Blue alone does not include it. OpenAI classifies GPT-6-Cyber as Critical for cybersecurity under its Preparedness Framework, compared with High for GPT-5.6-Cyber. That is OpenAI's own risk rating, not a Cody assessment.
Category comparison
These are provider-published specifications, not Cody benchmark scores. Follow the linked sources for current limits and endpoint-specific exceptions.
Pricing & comparisons
Set your usage. Your estimate updates as you type.
Example: 6,000 input tokens for 10 pages, plus a 500-token summary. Page lengths vary; adjust the numbers below.
One run sends your input to the model once and receives an answer. Tokens are pieces of text: input is what you send, output is the answer you receive.
GPT-6-Cyber
OpenAI
Estimated total (USD)
For the usage above · USD · API pricing, not a subscription
Estimates exclude taxes, tools, cache storage/writes, free allowances and custom discounts. Image estimates cover output only, not prompt or reference-image charges. Quality modes differ by model. Unlisted settings are not treated as free.
API and provider access
Availability
Approval required: available only to organizations with Daybreak Red access and a separate approval for GPT-6-Cyber, through the OpenAI API.
Daybreak Red organizations must require SSO and allowlist the IP addresses that send requests. OpenAI's regional-processing endpoints do not list GPT-6-Cyber.
Check live availabilityData and training
OpenAI says API data is not used for training by default. Trusted Access does not automatically include Zero Data Retention; default abuse-monitoring logs may be kept for up to 30 days, and projects with Zero Data Retention must set up Private Safety Processing.
This is a concise reading of the cited provider material, not legal advice. A third-party gateway can have different storage, routing, training, and residency terms from the model maker's direct API.
Read the provider policyFrequently asked questions
OpenAI's cybersecurity model for organizations approved for Daybreak Red, built to investigate vulnerabilities, reproduce them in authorized test environments and verify fixes. GPT-6-Cyber is a separate model in OpenAI's Daybreak program, not an access setting on a general model. OpenAI positions it for investigating vulnerabilities, reproducing them in an authorized test environment and verifying fixes. It reads text and images, returns text, and can use tools such as web and file search, code execution, computer use and MCP servers.
GPT-6-Cyber was released on November 5, 2026 according to the cited provider materials.
Approval required: available only to organizations with Daybreak Red access and a separate approval for GPT-6-Cyber, through the OpenAI API. The access routes listed in this guide are OpenAI.
$20 input · $100 output / 1M tokens. Rates for prompts up to 272K input tokens; longer prompts cost $40 input and $150 output per 1M tokens. Cached input costs $2 per 1M tokens on shorter prompts and $4 on longer ones. Cache writes cost $25 for 30 minutes or $50 for 12 hours. Batch is not supported.
Approval required: available only to organizations with Daybreak Red access and a separate approval for GPT-6-Cyber, through the OpenAI API. Daybreak Red organizations must require SSO and allowlist the IP addresses that send requests. OpenAI's regional-processing endpoints do not list GPT-6-Cyber.
OpenAI says API data is not used for training by default. Trusted Access does not automatically include Zero Data Retention; default abuse-monitoring logs may be kept for up to 30 days, and projects with Zero Data Retention must set up Private Safety Processing. The policy belongs to the provider route and account terms, so verify it again before production use.
One price per model, using the same input/output mix. We weight each API rate by its share of one million total tokens and use the lowest matching reviewed route. This is a comparison rate, not a cost per task or a quality score. Tokenizers, caching, reasoning and long context can change your actual bill.
75% input + 25% output · no cache discounts
Anthropic
OpenAI
Mistral AI
OpenAI
OpenAI
Anthropic
OpenAI
Anthropic
OpenAI
Anthropic
OpenAI
Meta
OpenAI
Anthropic
OpenAI
Anthropic
Anthropic
Anthropic
OpenAI
Anthropic
OpenAI
This model
Related comparisons
An OpenAI model for coding projects and assistants that carry out several steps with tools.
Open full comparisonOpenAI's approval-gated access to its flagship models for defensive security work, with fewer cybersecurity refusals.
Open full comparisonOpenAI's most capable model for difficult end-to-end work, combining frontier reasoning with a million-token context window and a broad set of agent tools.
Open full comparison