OpenAI has officially released GPT-5.6-Cyber, a specialized model built on the GPT-5.6 Sol architecture designed for offensive security and vulnerability research. Launched on August 10, 2026, the model demonstrates a massive performance leap in cybersecurity tasks, completing 95% of advanced prompts compared to just 1.5% for the standard Sol model with default guardrails. To manage the risks associated with such a powerful dual-use tool, OpenAI is restricting access through its new Daybreak program. This program is divided into two tiers: Daybreak Blue for defensive validation and Daybreak Red for vetted offensive research. Access is currently limited to major enterprise security partners, including IBM, CrowdStrike, and Palo Alto Networks. The model has already proven its efficacy by identifying two previously unknown vulnerabilities in the V8 JavaScript engine used by Google Chrome. This release marks a strategic shift for OpenAI, moving from broad safety guardrails to a controlled distribution model for high-stakes AI capabilities.