Unlimited text chats on the free tier of ChatGPT. That’s the headline change OpenAI announced today, and it lands at a moment when the company says the product recently crossed 1 billion weekly users.
The catch shows up fast. Unlimited applies to text, and only text.
OpenAI said separate limits still apply to files, images, voice and image generation. So the free account that now chats forever will still stop you cold when you try to feed it a PDF or ask it to draw something.
Driving the new free experience is GPT-5.6 Luna, which becomes the default model for Free and Go users and replaces GPT-5.5.
The Think button is the more interesting change
Free and Go users also get a new “Think” button, which lets them dial up higher reasoning power for complex questions. That’s reasoning-on-demand handed to people who aren’t paying anything, and historically that kind of capability has been the thing you upgrade for.
Paying customers get their own version of the same idea, with more granularity. Plus and Pro users are getting a thinking slider to adjust “how much” thought the model puts into an answer, tuned to the complexity and the number of steps involved in solving a query.
A button on the free tier, a slider on the paid one. The hierarchy is doing exactly what you’d expect it to do.
What Plus and Pro subscribers actually get
Plus and Pro users gain access to an upgraded GPT-5.6 Sol model, which OpenAI positions as better for quick tasks: questions, web research, advice, planning, writing and making decisions. The company said this new model will give more compact answers.
One detail worth pinning down, because the naming makes it easy to miss. This is a separate version from the GPT-5.6 Sol used for Codex and Work, which is unchanged.
The error numbers OpenAI is putting forward
OpenAI said an internal evaluation found that, compared with GPT-5.5-Instant, factual errors were 62% less common for GPT-5.6 Luna and 68% less common for GPT-5.6 Sol.
Internal evaluation is the operative phrase. These are OpenAI’s own numbers, measured against OpenAI’s own previous model, on a benchmark OpenAI didn’t detail. Fewer factual errors than GPT-5.5-Instant isn’t the same claim as few factual errors.
Still, a 62% reduction is the kind of figure that gets tested quickly once a billion people a week have their hands on it.
When you get it
The updated GPT-5.6 Sol is available to Plus and Pro users today. The other changes for Free and Go users are arriving this week.
Unlimited text chats and the new Think button for harder questions land next week. If you’re on the free tier, that’s the date that matters, and it’s the one to check your app against before you assume the cap is gone.
