OpenAI has begun rolling out its GPT-5.6 models to the public after the Trump administration lifted restrictions that had limited the release to government-approved entities. The flagship model, GPT-5.6 Sol, along with the lower-tier Terra and Luna variants, became available to all users on July 10.
What Changed After Government Testing
The rollout followed additional testing conducted by the Center for AI Standards and Innovation within the Department of Commerce. OpenAI sent technical experts to Washington D.C. who remained on-site to address questions during the evaluation process. The testing focused on the model’s capabilities and potential risks before clearing it for broad public access.
The Trump administration had originally pushed OpenAI to conduct a staggered release of GPT-5.6 in June, limiting initial access to approved entities. OpenAI said at the time that the staggered approach was not its preferred method and noted that both AI companies and the government were still working without finalized standards for model releases.
GPT-5.6 Sol: The Flagship Model
GPT-5.6 Sol represents OpenAI’s most capable model to date. It scored 94.1 on GPQA Diamond (beating Claude Fable 5’s 92.6 and Kimi K3’s 93.5), 73.0 on DeepSWE for coding tasks, and 88.8 on Terminal-Bench 2.1 for agentic coding. The model also introduces new “max” and “ultra” performance modes for handling complex reasoning and code tasks.
Terra sits below Sol as a mid-tier option, while Luna provides a lighter, more affordable alternative for everyday use. All three models support the same safety guardrails and API access tiers.
Anthropic’s Fable Restrictions Also Lifted
The government action against OpenAI was not an isolated case. In June, the Commerce Department banned foreign access to Anthropic’s Mythos and Fable models, essentially pulling them from the international market. The ban on Fable was lifted on June 30, with customer access restored the following day. The episode showed that the US government is actively negotiating case-by-case access to frontier AI models with the companies that build them.
What This Means for the AI Industry
The GPT-5.6 rollout underscores how AI release decisions are now intertwined with government oversight. Both OpenAI and Anthropic have had to adjust their deployment timelines based on federal feedback. The White House has maintained that no formal licensing or preclearance is required for AI model releases, but voluntary testing and meetings with government officials have become part of the process.
OpenAI also recently published usage data for its Codex agent platform, showing increasing adoption of AI coding agents across enterprise workflows. The company’s acquisition of NorthSlope, a deployment business, signals its push into enterprise infrastructure beyond just model access.
FAQ
What are the differences between GPT-5.6 Sol, Terra, and Luna?
Sol is the flagship model with the highest reasoning and coding performance. Terra is a mid-tier option that balances capability with cost. Luna is the lightest model, designed for everyday tasks and lower compute requirements.
Is GPT-5.6 available globally?
Yes, after the US government restrictions were lifted, GPT-5.6 Sol, Terra, and Luna are available to all users worldwide through the ChatGPT interface and the OpenAI API.
How does GPT-5.6 compare to Claude Fable 5?
GPT-5.6 Sol edges out Claude Fable 5 on several benchmarks: 94.1 vs 92.6 on GPQA Diamond, and 88.8 vs 88.0 on Terminal-Bench 2.1. However, Claude Fable 5 scored higher on the DeepSWE coding benchmark at 70.0 versus Sol’s 73.0.
Did the government actually approve the GPT-5.6 release?
A White House official stated that no formal permission is required or granted for AI model releases, and that timing decisions rest entirely with the companies. However, OpenAI conducted additional testing with Commerce Department officials before the broad launch.
