GPT-5.6 began rolling out to the public today, following a government mandated preview period that limited access to a small group of vetted partners for roughly two weeks. OpenAI’s three tier model family, Sol, Terra, and Luna, is opening across ChatGPT, the API, and Codex, closing out a rollout that faced closer government scrutiny than any AI model release before it.

For anyone catching up: Sol is the flagship model, built for the hardest tasks. Terra sits in the middle, matching the performance of the previous GPT-5.5 generation at roughly half the price. Luna is the lightweight option, meant for simple, high volume tasks where speed and cost matter more than raw capability. The full breakdown of what each tier is actually good for was covered here when the models first previewed on June 26.

Why It Was Gated in the First Place

Earlier this year, the Trump administration signed an executive order asking AI companies to voluntarily submit their most capable models for government review 30 days before public release. GPT-5.6 was the first major model family to go through that process in full. OpenAI initially limited access to a small group of government approved partners in late June, at the administration’s request, while the Commerce Department’s Center for AI Standards and Innovation ran additional testing.

At the time, OpenAI was clear that it did not see the process as something that should become permanent, but complied anyway, calling it the fastest path to getting the models into people’s hands responsibly. That review has now wrapped, moving faster than the original 30 day window suggested it might, with OpenAI sending technical staff to Washington to work through questions directly.

What Changed Today

As of today, GPT-5.6 began rolling out across ChatGPT, the OpenAI API, and Codex, OpenAI’s coding agent. OpenAI’s own announcement describes this as expanding preview access globally, language that leaves room for a staged rollout rather than instant access for every account at once.

That caution is holding up in practice. As of this writing, GPT-5.6 has not yet appeared in the model picker for every ChatGPT Plus account, with GPT-5.5 and earlier models still showing as the available options for some users. Anyone expecting to see it today shouldn’t assume access has failed if it isn’t there yet. A same day appearance isn’t guaranteed for every account, and checking back later today or tomorrow is the more likely path to seeing it show up.

Developers building on the API should note a separate detail. The older gpt-5.5-latest endpoint will not automatically point to the new models. Explicit model IDs, gpt-5.6-sol, gpt-5.6-terra, or gpt-5.6-luna, are required to actually use them. A new prompt caching system also went live alongside the release, offering steep discounts on repeated queries, which matters more for developers than everyday users but is worth knowing about for anyone paying for API access.

What This Means Going Forward

GPT-5.6’s rollout is part of a bigger shift happening across the industry this year. Frontier model launches are increasingly treated as events that require government sign off, not just ordinary product updates. The same week GPT-5.6 began opening to the public, xAI’s Grok 4.5 also went live, meaning every major AI lab now has a publicly available model at the same time for the first time in weeks. Only Google’s Gemini 3.5 Pro remains stuck in limited preview.

For everyday users, the practical takeaway is simpler than the politics around it. Real access to more frontier AI models than at almost any point this year is arriving, even if it’s arriving in stages rather than all at once. Picking between them is becoming less about waitlists and more about which one actually fits what you’re trying to do, once each one actually shows up in your account.

FAQ

Is GPT-5.6 available to everyone now?
It began rolling out today across ChatGPT, the API, and Codex, but OpenAI has described this as an expanding rollout rather than instant access everywhere. Some ChatGPT accounts are not yet seeing it in the model picker as of this writing, so a same day appearance isn’t guaranteed for every user.

Why was GPT-5.6 delayed in the first place?
It went through a government mandated review process tied to a 2026 executive order requiring advance testing of the most capable AI models before public release, the first major model family to complete that process in full.