Category: Technology News
OpenAI released GPT-6 Cyber on 5 November 2026, a security-focused version of its GPT-6 Astra model built for authorised vulnerability research. It is the first cyber-specific model OpenAI has rated Critical under its Preparedness Framework — the framework’s top tier for cyber risk.
The previous cyber model, GPT-5.6 Cyber, was rated High. GPT-6 Cyber is also the fourth cyber-specific model OpenAI has released this year.
What the model does
GPT-6 Cyber is tuned for security work such as finding vulnerabilities, writing proof-of-concept tests and helping draft patches. It is not a general chat model.
According to OpenAI, researchers using the model found 81 previously unknown high-severity vulnerabilities in the Firefox browser. In a controlled research pipeline, it reproduced 1,724 Linux kernel vulnerabilities and developed working exploits for 948 of them.
On OpenAI’s ExploitGym benchmark, which tests whether a model can exploit a specified vulnerability, GPT-6 Cyber scored 48.5%, against 32.1% for GPT-5.6 Cyber and 42.4% for GPT-6 Astra. OpenAI calls this a 51% relative improvement, or about 16 percentage points.
All of these benchmarks and counts are OpenAI’s own claims. No independent verification has been published.
Staying inside the authorised scope
One of the model’s key features is that it is built to stay inside the scope a researcher authorises. In OpenAI’s ExploitGym Honeypot test, which surrounds the model with supporting services that are deliberately easy to compromise, GPT-6 Cyber went out of bounds in 0.07% of graded runs without deployment safety instructions in place, against 5.51% for GPT-5.6 Cyber. GPT-6 Astra scored 0% on the same test.
Fouad Matin, head of OpenAI’s cyber preparedness and safety team, said the model does what it is asked over 99% of the time — again, OpenAI’s own result.
Who can use it
GPT-6 Cyber is not open to the general public. Access starts with a select group of approved Daybreak Red customers through OpenAI’s API and Codex, with IP allowlisting on by default. Daybreak Red is the tier of OpenAI’s Daybreak programme for offensive-style work such as penetration testing and red teaming. Verified defensive researchers use a separate Daybreak Blue programme.
OpenAI said GPT-6 Cyber will reach all Daybreak Red customers over the coming weeks, and GPT-5.6 Cyber is now opening to every customer in the tier.
The Critical rating is the reason for the gated access: a model that can build working exploits with little human help is a dual-use tool, useful to defenders and dangerous in the wrong hands.
Sources: OpenAI announcement, 5 November 2026; SiliconANGLE, 5 November 2026; Digit, 6 November 2026.
