OpenAI confirmed Wednesday that its next-generation GPT-5.6 family of models will launch Thursday, with CEO Sam Altman announcing the release on X. The lineup includes three distinct models: Sol, the flagship designed for deep reasoning and agent management; Terra, the balanced mid-range option offering a trade-off between capability and cost; and Luna, the fastest and cheapest to run. OpenAI says the models represent a meaningful improvement in cybersecurity capabilities while incorporating advanced safety guardrails, and that Sol significantly reduces factual errors compared to earlier versions.

The specific availability details remain unclear beyond Thursday's launch date. CNET reached out to OpenAI but did not receive an immediate response. (Disclosure: Ziff Davis, CNET's parent company, filed a lawsuit against OpenAI in 2025 alleging copyright infringement in training and operating its AI systems.) OpenAI first announced GPT-5.6 on June 26, but has deliberately slowed its rollout, similar to Anthropic's cautious approach with Claude Mythos 5. The company attributes this hesitation to the models' advanced capabilities, which make safety reviews more critical.

For the first time, the U.S. government is influencing release timing. On June 2, President Trump signed an executive order requiring the Defense Department to create a system for voluntary government review of new frontier models. Officials have 30 days to flag concerns, though engagement remains voluntary. OpenAI says it provided early access to the government and unspecified trusted partners, but declined to name them. A White House official clarified that the government did not formally approve the release—nor was that required—and emphasized the voluntary nature of the process. This marks a shift from the Trump administration's earlier stance that regulation would slow U.S. innovation in the global AI race against China.

Source: CNET