OpenAI previews GPT-5.6 Sol, Terra, and Luna models

OpenAI has initiated a limited preview of the GPT-5.6 series, introducing three new models: Sol as the flagship, Terra for balanced everyday work, and Luna for fast, affordable tasks. The company plans to make these models generally available in the coming weeks following this initial phase with trusted partners.

• GPT-5.6 Sol sets a new state of the art on Terminal-Bench 2.1 and achieves stronger results than GPT-5.5 on GeneBench v1 using fewer tokens.
• In cybersecurity, Sol is competitive with Mythos Preview on ExploitBench using approximately one-third of the output tokens and shows strong improvements on ExploitGym.
• The series includes a new max reasoning effort for Sol and an ultra mode that leverages subagents to accelerate complex work.
• OpenAI employs layered safeguards, including real-time classifiers and account-level signals, noting that Sol does not cross the Cyber Critical threshold under their Preparedness Framework.

The phased release aims to test safety measures against adversarial pressure while ensuring legitimate defensive work remains accessible. This approach allows OpenAI to refine safeguards and coordinate with the U.S. government before broader availability.