Updated on November 10, 2026
GPT-6 Cyber is a specialized OpenAI model built for vulnerability research, exploit development and patch verification in authorized environments, announced on November 5, 2026. It targets penetration testers, red teams and security researchers approved through the Daybreak Red program. API pricing sits at $20 per million input tokens and $100 per million output tokens, with no free tier, no trial and no ChatGPT access. OpenAI rates GPT-6 Cyber Critical for cybersecurity under its Preparedness Framework.
GPT-6 Cyber scores 48.5% on ExploitGym, the benchmark that measures whether a model can exploit a known security vulnerability on a remote target. GPT-5.6 Cyber reached 32.1% on the same test. In black-box web penetration testing, GPT-6 Cyber recalls 40.3% of vulnerabilities versus 32.0% for GPT-6 Astra. On RemoteExploitBench, it completed 95% of tasks against controlled GitLab, PostgreSQL and Jenkins deployments.
Beyond lab scores, OpenAI reports field campaigns on production-grade code. Researchers documented 81 previously unknown high-severity Firefox bugs (Mozilla has begun shipping fixes), built working privilege-escalation exploits for 948 Linux kernel vulnerabilities (588 of which were unknown), and flagged a critical RCE in Brotli (CVE-2026-111497, CVSS 9.2). All 304 critical Linux findings were disclosed to kernel maintainers with candidate patches. These figures come from OpenAI and have not yet been independently audited.
Daybreak is OpenAI's cyber program, launched in June 2026 with two tiers. Daybreak Blue serves defenders with flagship models that refuse fewer security-related queries. Daybreak Red unlocks specialist models for authorized offensive work: exploit development, proof-of-concept validation, red-team exercises. GPT-6 Cyber requires Red approval plus a separate model-level approval, both confirmed by your OpenAI account team.
Onboarding checks include identity verification, account security requirements, IP allowlisting for the API, dedicated monitoring and legal attestations. Critical infrastructure defenders can join a dedicated Daybreak Red group, with subsidized access through a $1 billion Daybreak for Frontline Defenders grant. Named partners at launch include Mozilla, Cisco, Palo Alto Networks, SentinelOne, CrowdStrike, CodeRabbit and Dragos (industrial security).
| API parameter | Value |
|---|---|
| Input (text + image) | $20 / 1M tokens |
| Cached input | $2 / 1M tokens |
| Output | $100 / 1M tokens |
| Context window | 1,050,000 tokens |
| Max output | 128,000 tokens |
| Model ID | gpt-6-cyber |
GPT-6 Cyber enters a growing field of gated cyber models. Google has announced a version of Gemini 4 Argon without cyber guardrails for trusted defenders in its Fairwind Program, and Anthropic reserves Claude Mythos Preview for Project Glasswing partners. All three programs share a common playbook: vetted cohorts, access verification, continuous monitoring. The broader developer tools market is adapting to this controlled-access pattern.
On alignment, OpenAI trained GPT-6 Cyber to stay within the researcher's authorized scope. In ExploitGym Honeypot tests, it avoided out-of-scope services 99.93% of the time without the deployment safety prompt and 100% with it. OpenAI has not detailed the remaining 0.07%. Starting November 29, 2026, content from Daybreak Blue sessions on personal accounts will feed model training, including abuse safeguards. Pricing and access terms may change at any time, so always verify on OpenAI's official site.
GPT-6 Cyber is a cybersecurity-specialized OpenAI model released on November 5, 2026. It is trained to find exploitable vulnerabilities, develop proofs of concept and verify fixes within authorized test environments. Access is limited to approved Daybreak Red customers through the API and Codex. OpenAI classifies it as Critical for cybersecurity under its Preparedness Framework, one level above the High rating of GPT-5.6 Cyber.
Daybreak Red is the offensive tier of OpenAI's cyber program, separate from the defensive Daybreak Blue tier. Applicants must pass identity verification, security reviews and legal attestations. GPT-6 Cyber also requires its own model-level approval on top of Red status. At launch, only a select group of Red customers received access; OpenAI plans to extend it to all approved Red members over the coming weeks.
On ExploitGym, GPT-6 Cyber scores 48.5% versus 32.1% for GPT-5.6 Cyber, a 51% relative improvement. Scope adherence jumps from 94.49% to 99.93% without a safety prompt (100% with one). The Preparedness rating rises from High to Critical. GPT-5.6 Cyber was built on GPT-5.6 Sol, while GPT-6 Cyber was released to select Red customers first.
The gpt-6-cyber endpoint charges $20 per million input tokens, $2 per million cached input tokens and $100 per million output tokens at standard short-context rates. The context window reaches 1,050,000 tokens with a 128,000-token max output. There is no free tier, no trial and no ChatGPT integration. Access requires Daybreak Red approval, confirmed by your OpenAI account team before the first request.
Verdict: An exploit pipeline that found 81 unknown Firefox bugs and built working attacks for 948 Linux kernel flaws in a single research campaign: GPT-6 Cyber fits security teams with formal authorization scopes, the infrastructure for controlled testing and the API budget to match. Smaller outfits or solo researchers without Daybreak Red clearance can start with GPT-6 Astra under Daybreak Blue for defensive code review.
« An all-in-one creative platform that brings together more than 30 image and video models in a single interface. It includes camera controls that let you direct like a real filmmaker, characters that remain consistent from scene to scene, and ready-to-use tools for ads and short-form content »
Claude Fable 5.2 
Claude Opus 5.5 
Gemini 4 Carbon 
