OpenAI Launches GPT-5.6 With US Government-Imposed Access Restrictions

OpenAI has unveiled its latest AI model series, GPT-5.6, but immediately limited access to a small group of trusted partners at the request of the US government. The company announced on June 28, 2026, that early access would only be granted to partners approved by the administration, emphasizing that such government-controlled access should not become the norm as it restricts valuable tools from developers, cybersecurity experts, and global collaborators.

The GPT-5.6 series includes three models: Sol, the flagship and most powerful model; Terra, a balanced model for everyday use; and Luna, a faster, more cost-effective option. The names reflect their relative scale, meaning sun, earth, and moon. Despite Sol’s advanced capabilities, the Trump administration has imposed launch restrictions on all three models as part of a broader effort to regulate advanced AI systems.

This move follows similar government pressure on other AI firms, such as Anthropic, which was ordered to block foreign access to its Fable 5 model and subsequently removed it entirely. The US government’s recent executive order requires AI companies to voluntarily submit their most advanced models for review 30 days before release, effectively creating a licensing regime that critics say could cause indefinite delays and hinder US competitiveness against China.

Dean Ball, a former White House AI advisor expected to join OpenAI, warned that unclear safety standards exacerbate the problem, potentially risking billions in AI infrastructure investments. OpenAI stated that GPT-5.6 availability will expand in the coming weeks and that it is collaborating with the government on a new cybersecurity executive order and a recurring review process for future model launches.

OpenAI highlighted Sol’s enhanced agentic abilities in coding, biology, and cybersecurity, featuring "max" and "ultra" modes that coordinate sub-agents to solve complex tasks. Sol reportedly outperforms Anthropic’s Mythos 5 in coding workflows while using only a third of the tokens. The model incorporates robust security measures designed to prioritize defensive cybersecurity over exploitation and integrates safety mechanisms directly into its core behavior rather than relying on external filters.

Initially, GPT-5.6 models are accessible only to select partners, but OpenAI plans to broaden access soon to ChatGPT, Codex, and API users. Pricing is tiered by model size, with Sol costing $5 per million input tokens and $30 per million output tokens, Terra at half that rate, and Luna at $1 and $6 respectively. Input tokens are user prompts, while output tokens are the model’s generated responses. OpenAI also improved prompt caching to reduce costs and improve predictability for repeated use.