OPENAI has introduced its new GPT-5.6 model lineup, featuring three tiers: Sol, Terra, and Luna. Sol is highlighted as the most powerful model specifically designed for cybersecurity tasks, capable of high-intensity reasoning. Following consultations with the U.S. government, the phased rollout aims to align with a recently established Executive Order concerning AI and national security. Terra offers competitive performance for everyday workload at reduced costs, while Luna serves as a budget-friendly alternative.
In cybersecurity tests, Sol matched performance with other systems but showcased a stronger focus on defensive tasks, such as identifying vulnerabilities rather than executing full cyberattacks. To mitigate dual-use risks, a multi-layer security architecture is implemented, including automated classifiers for monitoring behavior. The initial access is limited to selected partners, with plans for broader availability in the near future.