GPT-6 Cyber

GPT-6 Cyber (API id gpt-6-cyber) is OpenAI's Daybreak Red cybersecurity model for authorized vulnerability research, exploit validation, and security testing. OpenAI reports ExploitGym 48.5% (GPT-5.6 Cyber 32.1%), 40.3% recall on an internal web penetration testing eval (GPT-6 Astra 32.0%), and 95% on RemoteExploitBench, and classifies the model Critical in cybersecurity under its Preparedness Framework (GPT-5.6 Cyber: High). Official docs: 1.05M context window (922K max input, 128K max output), text+image input / text output, reasoning tokens, Chat Completions + Responses + Batch; list prices $20 input / $2 cached input / $25 cache writes (30 min) / $50 cache writes (12 hours) / $100 output per 1M tokens (short-context). OpenAI publishes no long-context (>272K input) rate for this model even though max input is 922K. Requires Daybreak Red access plus additional GPT-6-Cyber model approval; initially limited to a select group of approved Daybreak Red customers via the API and Codex, expanding to all Daybreak Red customers. GPT-5.6 Cyber remains available, and OpenAI is making it available to all Daybreak Red customers. available_in_zeroeval is false (gated).