OpenAI previews GPT-5.6 with Sol, Terra, and Luna models and stronger cyber safeguards

OpenAI has launched a limited preview of its GPT-5.6 series, introducing three new models: Sol, Terra, and Luna. Sol is positioned as the flagship, Terra is designed for everyday work with balanced features, and Luna offers fast and affordable access. While the lineup is new, Sol stands out with the most advanced safety features to date. Protections have been strengthened for higher-risk and sensitive activity, including repeated misuse, and additional preparedness evaluations are available to partners.

Building on this safety work, Sol is also OpenAI’s most capable model for cybersecurity. It moves the performance and efficiency frontier for demanding security tasks such as vulnerability research and exploitation. Sol also includes new technical advances: an ultra mode that leverages subagents for complex tasks and a max reasoning effort allowing deeper problem-solving. Terra offers performance close to GPT-5.5 at half the cost, while Luna brings strong capability at OpenAI's lowest price point.

These models share updated safeguards tailored to each model's capabilities. GPT-5.6 actively refuses prohibited cyber assistance, including evasion attempts, and uses real-time misuse classifiers. For higher-risk cases, output may be escalated for review by a larger model and withheld if assessed as noncompliant. Initial preview access is limited to select partners through the API and Codex, with plans for broader rollout soon.