GPT-6-Cyber

Cybersecurity model in the GPT-6 line, added to OpenAI's API changelog on 2026-11-05 ("Use GPT-6-Cyber to investigate vulnerabilities, reproduce them in an authorized test environment, and verify fixes"). ACCESS IS GATED, so the published rate is not self-serve: the model page states "Your organization needs Daybreak Red access and approval to use GPT-6-Cyber. Your account team can confirm access", and the pricing page files it under 'Cyber - Trusted Access: Daybreak'. Text+image input, text output; 1,050,000-token context with a stated maximum of 922,000 input tokens, 128,000 max output, reasoning tokens supported. No knowledge cutoff is published on the model page, so that field is null. Standard prices per 1M for requests with up to 272K input tokens, which is the band the price fields carry: $20.00 in / $2.00 cached in / $100.00 out; cache writes cost $25.00 for 30-minute retention and $50.00 for 12-hour retention. Unlike gpt-5.6-cyber, a long-context band IS published. Chat Completions and Responses are supported; Batch is not. Snapshot list is the bare alias gpt-6-cyber only. STATUS: recorded as ga on the same reading as the gpt-5-6-cyber row - a released, priced, access-gated model with no preview wording from OpenAI; the gate is recorded here in prose because the schema has no field for it. Lifecycle fields are null: OpenAI states no deprecation for this model. The same day OpenAI deprecated gpt-5.5-cyber-preview (shutdown February 1, 2027, replacement "the most capable cyber model available to you"), which this catalog does not carry. Recorded 2026-11-10, five days after OpenAI's date: the event is theirs, the delay is ours.

Source: OpenAI official documentation ↗

Explore more

GPT-6-Cyber — questions & answers