OpenAI is changing what happens when a small-business owner opens ChatGPT and starts typing. Two updates arrived close together, and they point in opposite directions on purpose. Free users are getting a better model handed to them automatically, with nothing to set up. Paid users are getting a manual control over how hard the model thinks before it answers.
If you run a business and you have felt behind on all of this, that split is good news. One change means the free tool you already use got more capable on its own. The other change surfaces a skill worth learning: knowing when a fast answer is fine and when a hard question deserves more thought.
Here is what actually changed and how to use it this week.
What free users get: a smarter default and unlimited text
OpenAI is making GPT-5.6 Luna the new default model for its Free and Go users. Luna is the fastest and most affordable model in the GPT-5.6 family, giving free users access to OpenAI's newest model generation without touching a single setting.
Two more changes are coming for free users starting next week:
- Unlimited text chats. Limits will still apply to file uploads, images, and other tools.
- A new Think button. For harder questions, users can tap Think to give GPT-5.6 Luna more time to reason through the answer.
That last one is the part worth slowing down on. For years, using AI well meant knowing which paid tier to buy. Now a free user can decide, question by question, whether a task deserves a fast reply or a more careful one. That decision is the whole game.
What paid users get: a slider for reasoning effort
For Plus and Pro subscribers, OpenAI updated GPT-5.6 Sol, its flagship model, to be more reliable with facts and to give more focused answers. A new reasoning slider lets users choose how much thought ChatGPT puts into a response, with the available settings depending on the plan. The slider is rolling out across web, mobile, and desktop.
Plus users get:
- Instant for fast everyday responses
- Medium for standard reasoning with GPT-5.6 Sol
- High for extended reasoning with GPT-5.6 Sol
Pro users also get:
- Extra High for the highest reasoning effort available with GPT-5.6 Sol
- Pro for GPT-5.6 Sol Pro, built for difficult tasks and longer-running work
Developers building on the API see an even finer set of controls: six reasoning-effort levels, from none up through low, medium, high, xhigh, and max.
Notice the shape of it. The free tool gives you one button (think harder, or do not). The paid tool gives you a dial. And the developer tool gives you six notches. The same idea, reasoning effort, shows up in three sizes depending on how much control you want.
Why this counts for a busy owner
Reasoning effort is a real trade. A fast answer arrives quicker. Higher reasoning effort gives the model more room to work through difficult problems, although it still does not guarantee a correct answer. In an OpenAI internal evaluation of financial, medical, and legal prompts requiring factual detail, responses containing at least one factual error were about 62% less common with GPT-5.6 Luna and 68% less common with GPT-5.6 Sol than with GPT-5.5 Instant.
Fewer responses with at least one factual error on financial, medical, and legal prompts, comparing GPT-5.6 Luna and Sol against GPT-5.5 Instant.
Source: OpenAI internal evaluation, 2026So the practical rule for your team is simple:
- For everyday chat (a quick email draft, a rewrite, a brainstorm), the default is good enough now. Let it run fast.
- For a question where being wrong carries real consequences, such as a contract clause, tax issue, dosage question, or compliance requirement, use more reasoning and verify the answer against a trusted source or qualified professional.
You do not need a bigger plan to make that call. You need the judgment to know which kind of question you are asking. That judgment is teachable, and it is worth teaching to whoever on your team touches these tools.
You do not need a bigger plan to make that call. You need the judgment to know which kind of question you are asking.
Where this sits in the 7 Levels of AI Proficiency
At LaunchReady we measure this kind of skill with The 7 Levels of AI Proficiency. The early levels are about getting comfortable typing into the tool and trusting a first answer. Somewhere in the middle, an operator learns to check the answer instead of accepting it, and to match the effort to the stakes of the question.
The Think button and the slider are the mainstream arrival of that exact idea. The model got better. The setting for how hard it thinks is now sitting right there in the interface. But the model cannot tell you that a contract clause is high-stakes and a birthday-card draft is not. That read stays with the human, and it is the difference between someone who uses AI and someone who runs it well.
Your next step
Once the new controls reach your account, run one small test. Ask it an everyday question and let it answer fast. Then ask it a real business question where being wrong would cost you, and this time tap Think (free) or move the slider up (paid). Read both answers side by side. That five-minute exercise teaches your team the one habit these updates reward: spending the extra thinking only where it earns its keep.
Related reading: Level 3: The Lieutenant (Critical Thinker).
Sources
- Improving GPT-5.6 Sol in ChatGPT
- Introducing GPT-5.6
- GPT-5.6 in ChatGPT (Help Center)
- Reasoning guide (OpenAI API docs)
- OpenAI launches its new family of models with GPT-5.6 (TechCrunch)
Frequently Asked Questions
Do I have to change any settings to get the smarter free model?
No. GPT-5.6 Luna becomes the default for Free and Go users, so the upgrade happens on its own. The Think button is there when you want to push a specific answer harder.
Is the reasoning slider available to everyone?
No. It is rolling out to eligible paid plans, including Plus and Pro. The options depend on the plan. Free and Go users instead receive a Think button powered by GPT-5.6 Luna.
What are the models in the GPT-5.6 family?
Three tiers: Sol, the flagship; Terra, a lower-cost model with performance competitive with GPT-5.5; and Luna, the fastest and most affordable. The family launched on July 9, 2026.
When did the free-tier expansion roll out?
OpenAI's own post describes the rollout in relative terms ("this week" and "next week") and appears to be phased rather than switched on everywhere at once. Check your own ChatGPT app to see whether the new default and the Think button have reached your account yet.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.