OpenAI is about to drop GPT-5.6 with a 1.5 million token context window and it could arrive any day

Turns out the prediction markets called it. OpenAI's GPT-5.6 didn't just "drop", it landed on 26 June, wrapped in more government paperwork than a passport renewal.

Here's the twist nobody priced in: you probably can't use it. Not because it's not ready, but because the Trump administration asked OpenAI to hold it back.

GPT-5.6 actually arrived as three models, not one. Sol is the flagship, built for coding, science and long agentic tasks. Terra is the everyday workhorse, matching GPT-5.5 for half the price. Luna is the cheap, fast option for high volume jobs. There's a naming shift too: the number now tracks the generation, while Sol, Terra and Luna are capability tiers that can update on their own schedule from here.

The reason for the government involvement is blunt. GPT-5.6 Sol is OpenAI's strongest model yet at finding and exploiting software vulnerabilities, the kind of cyber capability that makes national security types nervous. OpenAI says Sol didn't cross the "critical" risk threshold in its own testing, it found bugs in Chromium and Firefox but couldn't chain them into a full working exploit, but Washington still wanted a say before anyone got broad access. Anthropic went through the same process with its Fable 5 and Mythos 5 models, so this is now standard practice for frontier releases, not a one off.

For now, access sits with about 20 vetted partners, with OpenAI expecting to widen that group next week and reach general availability in the coming weeks. Axios reckons mid July is the realistic target for the rest of us.

As for that 1.5 million token context window everyone was whispering about before launch, OpenAI hasn't confirmed it in the official rollout, so file it under plausible but not yet locked in until the broader release lands with full specs.

What's actually interesting here isn't the model, it's the precedent. The most powerful AI systems in the country are now getting a government sign off before the public sees them, and both OpenAI and Anthropic have quietly accepted that as the cost of shipping. That's a bigger story than a bump in context window, and it's the one worth watching over the next few weeks.

If you're waiting on ChatGPT to hand you Sol, don't hold your breath, hold your refresh button. Mid July, if the government agrees.