Daybreak Red

OpenAI's Daybreak Red tier gives vetted security firms access to GPT-5.6-Cyber, a specialized model built on GPT-5.6 Sol and trained to find zero-day vulnerabilities, build exploit chains, and refuse less often on high-risk dual-use tasks. According to OpenAI's internal Advanced Cybersecurity Completion Rate (ACCR), GPT-5.6-Cyber completes 95.0% of approved advanced-cyber requests, compared with roughly 1.5% for GPT-5.6 Sol under default safeguards and 2.0% via Daybreak Blue. The predecessor GPT-5.5-Cyber scored 57.3% on the same internal metric, a jump OpenAI attributes to researcher feedback on refusals. On third-party-style benchmarks reported by OpenAI, GPT-5.6-Cyber beats both GPT-5.6 Sol and GPT-5.5-Cyber on ExploitGym (turning known vulnerabilities into working exploits) and on discovering and calibrating novel zero-days. However, on a vulnerability-report-writing test it trails GPT-5.6 Sol because the specialized model tends to produce shorter, less detailed reports. Coverage notes GPT-5.6 Sol via Daybreak Blue retains an edge on ExploitBench against Google's V8 engine with defensive protections enabled, illustrating the tier's different operating points.

OpenAI announced an expansion of its Daybreak program on August 10, 2026, splitting access into two tiers: Daybreak Blue (general-purpose GPT-5.6 Sol without standard security-prompt guardrails) and Daybreak Red (restricted access to GPT-5.6-Cyber, a purpose-trained cybersecurity model built on GPT-5.6 Sol). Daybreak Red is positioned for authorized vulnerability research, exploit validation, and security testing by approved defenders, targeting a narrowing window before AI-enabled offensive threats scale. The tier-based design separates vetted defensive cyber work from general API consumers. GPT-5.6-Cyber is explicitly trained to improve performance on specialized cybersecurity tasks such as finding zero-day vulnerabilities and developing exploit chains, and to reduce refusals on certain higher-risk, dual-use cyber tasks. OpenAI states it delivers state-of-the-art performance on cybersecurity benchmarks and deploys system-level safeguards to screen cyber-related requests in production, but Daybreak Blue and Red tiers remove or reconfigure those guardrails so approved defenders can complete authorized work. Access remains gated to vetted security partners.