OpenAI has announced a limited initial rollout of its new GPT-5.6 models, Sol, Terra, and Luna, to trusted partners, with a broader release planned in the coming weeks. The decision follows a request from the US government, which was briefed on the models’ capabilities ahead of the launch. The government asked for a limited release while federal authorities develop a framework for evaluating advanced AI models under a recent cybersecurity executive order.
OpenAI intends to make the models available across ChatGPT, the API, and Codex in the coming weeks. However, the company emphasized that government preview requirements should not become a permanent standard for future frontier AI releases. The launch comes amid increasing scrutiny of advanced AI systems in the United States.
The GPT-5.6 models introduce advances in reasoning, coding, and cybersecurity capabilities. OpenAI described GPT-5.6 Sol as its flagship model, designed for demanding scientific workloads. Terra is positioned as a balanced model for enterprise and developer tasks, offering performance comparable to GPT-5.5 at roughly half the cost.
Luna is the company’s fastest and most affordable model for cost-sensitive applications. Sol will be available on Cerebras hardware in July for select customers, delivering inference speeds of up to 750 tokens per second. The model features Max Reasoning Effort, allowing it to spend more time solving complex problems, and an Ultra Mode that deploys multiple specialized sub-agents to improve coding, research, and multi-step workflows.
Meanwhile, US-based AI company Anthropic has received approval from US authorities to restore limited access to its Mythos 5 model for certain trusted partners after addressing national security concerns.
