OpenAI has already announced GPT-5.6, but most people still cannot use it. The new model family is currently in limited preview, and OpenAI has not confirmed an exact public launch date for ChatGPT, Codex, or the API.

OpenAI says GPT-5.6 comes in three models: Sol, Terra, and Luna. Sol is the flagship model for frontier reasoning and long-horizon agentic work. Terra is the balanced model, while Luna is the fastest and cheapest option for high-volume tasks.

OpenAI confirmed on June 26 that it is previewing the GPT-5.6 family with select trusted partners and organizations. The company says broader access for ChatGPT, Codex, and API users is coming soon.

TechCrunch reported that OpenAI limited GPT-5.6 access to a small group of trusted partners after a U.S. government request. Reuters also reported that the Trump administration asked OpenAI to delay the wide release of GPT-5.6.

OpenAI has not said the delay is permanent. Its official language still points to a broader rollout “soon,” but anything more specific than that remains unofficial.

Polymarket’s release-date market currently has July 9 as the leading outcome, followed by July 14. Earlier market activity had pushed July 7 much higher, but the odds have moved as traders react to the lack of an official announcement.

That does not mean GPT-5.6 will definitely launch this week. Prediction markets are useful for tracking sentiment, but they are not launch confirmations. The only confirmed timeline from OpenAI is that the preview will expand after more testing.

OpenAI may open GPT-5.6 first to paying ChatGPT users, then developers through the API, or it could prioritize Codex and enterprise access before a wider consumer rollout.

The pricing also shows that OpenAI is no longer treating one model as the answer for every use case. Sol is for the hardest reasoning and agentic work. Terra looks like the everyday model for developers who want strong performance at a lower cost. Luna is clearly built for speed and scale.

OpenAI is also changing prompt caching with GPT-5.6. Cache writes are billed at 1.25x the model’s uncached input rate, while cache reads still get a 90% cached-input discount. OpenAI says the new system supports explicit cache breakpoints and a 30-minute minimum cache life.

GPT-5.6 Sol is being pitched as a stronger model for coding, scientific reasoning, cybersecurity, long-horizon planning, and agentic workflows. OpenAI says it improves reliability and efficiency across demanding real-world tasks, which is exactly where competition with Anthropic, Google, and Meta is getting more intense.

The public launch now depends on how quickly OpenAI clears the remaining review and rollout work.

A release this week is possible, and prediction markets clearly expect movement soon. OpenAI has still not confirmed July 7, July 9, or any other date.

For now, GPT-5.6 is real, the limited preview is live, and the public launch looks close. The next OpenAI announcement should finally answer the only question most users care about: when they can actually use it.

Note: Links to online stores in articles can be so-called affiliate links. If you buy through this link, MIXED receives a commission from the provider. For you the price does not change.