OpenAI is rolling out a revised GPT-5.6 experience in ChatGPT that gives eligible paid users direct control over reasoning effort. At the same time, GPT-5.6 Luna is becoming the default model for Free and Go users.
The distinction matters: free users are not getting OpenAI’s flagship Sol model. They are getting the faster, lower-cost Luna model for regular text chats and, starting next week, a Think button for harder questions.
Paid users can choose how much ChatGPT reasons
The new slider lets Plus and Pro users move between quicker answers and deeper reasoning without picking from a long list of model names.
The lowest Instant setting still runs on GPT-5.5 Instant. Medium and High use GPT-5.6 Sol, while Extra High is available on eligible higher-tier plans. The Pro setting uses GPT-5.6 Sol Pro for difficult or longer-running tasks.
ChatGPT can also switch automatically from Instant to Medium when a question needs more reasoning. OpenAI says those automatic switches do not use the separate allowance for manually selected reasoning.
That setup is clearer than asking people to understand every model tier. A quick question can stay on Instant, while research, planning, writing, coding, or a complicated decision can get more reasoning through the same control.
Free users get Luna, not Sol
GPT-5.6 Luna is rolling out as the default model for Free and Go users. OpenAI describes Luna as the fastest and most affordable member of the GPT-5.6 family, below the balanced Terra model and flagship Sol.
9to5Mac reports that OpenAI is offering unlimited Luna-backed text chats to free users. OpenAI’s current help page uses more careful language: it says free-tier access to GPT-5.5 Instant remains limited within a five-hour window and that limits can change based on demand, safeguards, and individual usage.
In other words, “unlimited” does not mean unlimited access to Sol or GPT-5.5 Instant. It refers to basic text chat continuing on Luna, while premium models and other features keep their own allowances.
Free and Go users will also get a Think button powered by Luna. It gives the model more time for a difficult question without moving the user to Sol. OpenAI says that rollout starts next week.
Work, Codex, and Apple Intelligence are separate
This update is for standard ChatGPT conversations. OpenAI says model availability in ChatGPT Work, Codex, and the API is unchanged, so existing plan limits and model choices still apply there.
The announcement also does not describe a change to the separate ChatGPT extension used by Siri and Apple Intelligence on iPhone. For now, the practical difference is inside ChatGPT itself: paid users get a simpler reasoning control, while free users move to Luna for everyday chat.
