June 26th they said. Wednesday they teased it. Today it is actually here.
The whole family has landed. Sol. Terra. Luna.
They are available now in ChatGPT. In Codex. In the API. If you aren’t seeing them yet, wait 24 hours. That’s the rollout window. Global spread takes a minute.
OpenAI hasn’t exactly lost its grip. Nearly four years after that first ChatGPT drop, they’re still the biggest name in town. Even with Google circling. Even with Anthropic getting clever. The lead is shrinking maybe but it is still there.
A quick side note because the universe loves irony. CNET’s parent company, Ziff Davis. They sued OpenAI in 2025. Claimed copyright theft on training data. Just saying.
Three New Faces
Sol is the heavyweight.
It handles the heavy lifting. Complex reasoning. Technical execution. Design work that usually breaks lesser models. OpenAI is calling it their best coding engine yet. They say it beats Anthropic’s Fable 5 on the Artificial Analysis Coding Agent Index. Not just smarter. Cheaper too. Fewer tokens. Less time. Less money.
Then there is Terra.
The middle child. The reliable one. You’ll probably use this for everyday stuff. It is the workhorse for most people’s workflows. Not flashy. Just done.
And Luna?
Fast. Cheap. It prioritizes speed and efficiency over deep thinking.
The design chops got an upgrade across the board. ChatGPT-5.6 can look at a rendered result. It can see visual issues. It doesn’t just guess it looks right. It checks.
Also security.
They call it their strongest cybersecurity model yet. High performance. Low token cost. Which is ironic considering why it wasn’t public earlier.
Remember the wait?
It wasn’t just marketing delay. Frontier models like GPT-5.6. Anthropic’s Mythos 5 and Fable. The cyber capabilities scared everyone a bit. Governments got involved. Anthropic let in a few insiders first. Then when they tried to open up. The Trump administration stepped in. Ordered a ban on foreign nationals. Effectively killed the launch while safety folks reviewed it.
OpenAI did the same. Staggered release. Professionals first. Tech workers. Security teams. Then you.
Safety precautions can feel like speed bumps when you just want to see what the AI can do.
Work Does The Work
Here is the other thing.
It wasn’t just about the new brain. It was about the hands too.
Enter ChatGPT Work.
It’s an agent. Lives inside ChatGPT. It goes out. Touches your apps. Reads your files.
It doesn’t wait for you to tell it every tiny step. Powered by the new 5.0 architecture. It reasons. It plans. It executes.
You sit back. Maybe it runs Scheduled Tasks while you’re at the gym or in a meeting or ignoring your computer entirely. It gathers info across your digital life and builds a plan.
Is it convenient?
Probably.
Is it a little terrifying to have an autonomous agent messing with your files without much interaction?
You decide.
The rollout is complete. Or nearly. Sol is thinking hard. Luna is moving fast. And now Work is doing stuff in the background.
We will see where this leads.
