OpenAI to Publicly Launch GPT-5.6 Sol, Terra and Luna Models
OpenAI will launch its GPT-5.6 Sol, Terra, and Luna models on Thursday, July 9, following the removal of a U.S. government restriction. CEO Sam Altman confirmed the release via X on Wednesday, July 8, after the White House granted approval for the broader rollout of the agentic AI suite.
The Tiered Architecture of GPT-5.6
OpenAI is positioning the 5.6 release as a diversified product line rather than a single model.
- Sol: The flagship model. OpenAI identifies this as its strongest model yet in terms of agentic capabilities in coding, biology and cybersecurity.
- Terra: A balanced model designed for general productivity. OpenAI claims Terra performs similarly to GPT-5.5 but operates at half the cost.
- Luna: The entry-level, high-speed model. This version is marketed as the most affordable option in the company’s current catalog.
White House Intervention and National Security Directives
Axios reported Wednesday that the broader launch occurred only after OpenAI received explicit approval from the White House. On June 26, OpenAI disclosed that it had previewed the GPT-5.6 capabilities with the government and subsequently limited the release at the government's request.
The restriction involved a tight circle of “trusted partners” who maintained access while the general public was locked out. Politico reported on June 25 that the pause was the result of consultations between OpenAI, the Office of Science and Technology Policy, and the Office of the National Cyber Director.
Competitive Parallel: The Anthropic Export Control Precedent
According to a June 30 announcement from Anthropic, the company began restoring access to those models only after the federal government lifted specific export controls.
Anthropic had originally introduced Fable 5 and Mythos 5 on June 9, asserting that it had built-in safeguards to prevent misuse. However, days later, the company disabled access in response to a government directive citing “national security authorities.”