OpenAI ships GPT-5.6-Cyber behind a vetted Daybreak "Red" tier, splitting its defender programme in two

On 10 August 2026 OpenAI launched GPT-5.6-Cyber, a model built on GPT-5.6 Sol and trained for vulnerability research and exploit-chain work, and split its Daybreak defender programme into two tiers: Blue for approved defenders using the general frontier models with adjusted safeguards, and Red, an applicant-vetted tier that is the only way to reach the cyber model. OpenAI reports the model completes 95.0% of its internal advanced-cybersecurity prompts where standard Sol completes 1.5%, and says its researchers used it to find two previously unknown Chrome V8 vulnerabilities that chain into a sandbox escape, disclosed to Google as CVE-2026-15903. Listed pricing is $12.50 per million input tokens and $75 per million output.

What it means The frontier labs are now shipping deliberately less-refusing models to a gated list of defenders — so "can the vendor even sell me this capability" becomes a procurement question, not just a technical one.