OpenAI releases GPT-5.6 to everyone as US ban lifts
The company is publicly rolling out the GPT-5.6 Sol, Terra and Luna family nearly two weeks after access was limited to a small circle of partners at the US government's request for a security review.
- OpenAI began publicly releasing the GPT-5.6 family on July 9, after nearly two weeks in which access was restricted to a small circle of trusted partners.
- The flagship model is GPT-5.6 Sol, alongside the cheaper Terra and Luna versions.
- The limits were lifted following a US national security review.
- On the same day, Meta unveiled its Muse image-generation model, while Microsoft began shifting some queries to its own MAI family.
OpenAI began publicly releasing its next-generation models, called GPT-5.6, on Thursday, July 9, ending a nearly two-week period in which access had been limited to a narrow group of trusted partners. The restriction was put in place at the US government’s request for a security review, and lifting it means the tool becomes available to the wider public as of today.
The family consists of three models. The flagship is GPT-5.6 Sol, which the company describes as its most capable system to date, while Terra and Luna are the cheaper, more cost-efficient versions aimed at broad use. The original release date was pushed back last month over concerns about misuse risk, specifically in cyberattacks, coding and biology.
Why did the government ask for a pause?
The delay was not technical but the result of a security assessment. According to available information, the US administration requested that early access remain narrow until a national security review was complete, amid worries that a more advanced model could make harmful applications easier. Only after that review were the restrictions formally lifted, so OpenAI can now go public without the government-requested access conditions.
The case shows how far releases of the most advanced artificial intelligence models have moved beyond being purely business events. When a government becomes part of the launch process for a commercial product, the line between technological development and security policy grows thinner. It is worth staying cautious about exactly what the model can do, since the official descriptions of its capabilities come from the company itself and have not yet been independently verified in wide use.
The same day, moves by rivals
The GPT-5.6 launch coincided with several market moves. Meta unveiled an image-generation model called Muse, which it says will be integrated into Instagram, WhatsApp and Meta AI. That positions the largest social network more clearly in the race over generative image tools, in a space where it already has an enormous user base.
Microsoft, meanwhile, began redirecting tens of thousands of weekly queries away from third-party models and onto its own MAI family, inside tools such as Excel and Outlook. The motive is cost reduction, since relying on outside models is expensive, and a company building its own systems can ease its budget over the long term. All three moves in a single day show that the artificial intelligence landscape is currently being reshaped quickly and on several fronts at once.
For users in the region, the practical difference will not be immediate. Availability, pricing and language support for Serbian and neighboring languages usually arrive in waves after a global release, so the real effect is realistically expected only after a few weeks of use in real conditions.
Are you more interested in what the new model can do, or in the fact that a government was able to delay its release?
