Skip to content
TrustList
All launches
AI model launch

GPT-6-Cyber

OpenAI's restricted security model from 5 November 2026, for investigating vulnerabilities and verifying fixes: a 1,050,000-token context at $20 in and $100 out per million tokens, for Daybreak Red organisations with further approval.

Launched 2026-11-05AI model

About this launch

GPT-6-Cyber is a security model that OpenAI added to its API on 5 November 2026 under the model code gpt-6-cyber. OpenAI gives its purpose as investigating a vulnerability, reproducing it in an authorised test environment, or verifying a fix. It takes text and images and returns text.

## Closed to almost everyone at launch

An organisation cannot simply select this model. It needs Daybreak Red access, OpenAI's highest cybersecurity access programme, plus additional approval for this model, and requests must set the access value daybreak_red. OpenAI requires Daybreak Red organisations to use single sign-on and to enforce multi-factor authentication through their identity provider. FourWeekMBA reports OpenAI as saying the model is first available through the API and Codex to a select group of approved customers, with all Daybreak Red customers to follow.

## What OpenAI claims for it

OpenAI's figure, as reported by FourWeekMBA, is 48.5% on ExploitGym, a test of whether a model can exploit a specified vulnerability in a remote target, up from 32.1% for GPT-5.6 Cyber. The same report says OpenAI rates the model Critical in cybersecurity under its Preparedness Framework, up from High for GPT-5.6 Cyber. These are the maker's own results, so we store no score for them.

## Priced well above the general GPT-6 models

Input costs $20 and output $100 per million tokens at standard short-context rates, with cached input at $2. Cache writes cost $25 per million tokens for 30 minutes or $50 for 12 hours. The Chat Completions and Responses endpoints are supported, and Batch is not. GPT-5.5-Cyber Preview retires on 1 February 2027, so anyone still using it has to move before then.

## Key facts

**Maker:** OpenAI **Released:** 5 November 2026 **Model code:** gpt-6-cyber **Context window:** 1,050,000 tokens **Maximum input:** 922,000 tokens **Maximum output:** 128,000 tokens **Input and output:** text and image in, text out **Price per million tokens:** $20 input, $2 cached input, $100 output **Availability:** API, Daybreak Red organisations with additional model approval

Not yet independently verified: Not yet independently verified. The specifications and prices are from OpenAI's own model page, changelog and access guide, and the ExploitGym result and risk rating are OpenAI's own as reported by FourWeekMBA on 5 November 2026; we found no independent test. We will update this when it can be confirmed, and remove this note.

Discussion

Questions & answers about GPT-6-Cyber.

No questions yet. Be the first to ask about GPT-6-Cyber.