OpenAI introduced GPT-6 Cyber on November 5. Initial API and Codex access is limited to selected approved Daybreak Red customers; broader Red access is planned. GPT-5.6 Cyber is being opened to all Red customers.

Vendor evaluations

OpenAI reports 48.5% on ExploitGym versus 32.1% for GPT-5.6 Cyber. These vendor results were not independently reproduced.

The Linux model is now named

The announcement identifies GPT-6 Cyber as the model behind the earlier 948 Linux privilege-escalation exploits. Candidate patches still require review. This does not confirm a connection to unannounced Astra usage labels.

Listed API pricing

The exact model page lists Standard short-context rates of $20 input, $100 output and $2 cached input per million tokens. Separate long-context and Batch estimates are not established here. Daybreak Red access and model approval are required.