If you use the free version of ChatGPT, it is about to look and feel different. OpenAI is rolling out GPT-6 to all ChatGPT users, along with a new feature called Intelligent UI that lets the chatbot answer with charts, buttons, forms and small interactive tools instead of just paragraphs of text.
Paying subscribers started getting the update on Wednesday, Oct. 7. Free users and those on the lower-cost Go plan get it starting today, Oct. 8. OpenAI says ChatGPT now has more than 1.2 billion weekly users, so this is one of the biggest changes to the app in a while.
Here is what is changing, what OpenAI is promising, and what it has not said yet.
Which Version Do You Get?
OpenAI released its first GPT-6 models to paying customers last month. This rollout brings two versions, both tuned for everyday conversation, into the main Chat tab:
- GPT-6 Sol for Plus, Pro, Business and Enterprise subscribers. Enterprise users get it once their company’s administrators turn it on.
- GPT-6 Luna for Free and Go users.
OpenAI has not published a head-to-head comparison of Sol and Luna, but the split follows a familiar pattern: paying customers get the more capable model. The update only affects Chat. The models behind ChatGPT’s Work mode and its Codex coding tool are not changing.
What Is Intelligent UI?
Until now, ChatGPT has mostly answered in text, with the occasional image or table. “ChatGPT has predominantly been a text-based interface,” OpenAI product manager Aarush Selvan told TechCrunch. The most helpful answers, he said, “aren’t just text.”
With Intelligent UI, ChatGPT decides what format suits your question. A comparison might appear side by side. An explanation might come with a diagram you can interact with. A simple question still gets a plain text answer.
Examples OpenAI Has Shown
- Everyday planning: a timeline for cooking a Sunday roast, a road trip map with stops, or a visual plan for organizing a wardrobe.
- Learning: interactive explainers on topics like GDP, the Monty Hall problem or how an airplane wing creates lift, where you can change inputs and see the result change.
- Quick tools on demand: a bill splitter, a retirement savings calculator or a simple game, built inside the conversation.
Behind the scenes, OpenAI says it built a library of ready-made interface pieces and a system that draws the interface on screen while the model is still writing its answer. It says it trained the model to judge these interfaces on clarity, usefulness and completeness, but it also admits that the model’s design judgment and the range of what it can build “still need work.”
TechCrunch saw a demo in which a product manager asked ChatGPT to explore how an airplane wing generates lift, and it quickly produced several diagrams showing how planes stay in the air. OpenAI has also shown visuals for recipes, a diagram of how a bicycle works, a map for a multiday hiking trip and a personal savings calculator.
If you find the visuals distracting, TechCrunch reports you can dial them down in ChatGPT’s settings, similar to how you can adjust its personality.
Faster Answers, Smarter Searching
The other big change is speed. Earlier reasoning models would think through a problem completely before showing you anything. GPT-6 can start answering while it is still reasoning, adding to its reply as it goes.
OpenAI gives a few numbers from its own internal tests:
- On questions that need a web search, the fast GPT-6 Instant mode starts answering 44% sooner on average than the previous GPT-5.6 Instant.
- On everyday multistep tasks, the most thorough GPT-6 setting starts answering about as quickly as a mid-level GPT-5.6 setting and scores better overall than GPT-5.6 at its most thorough.
- OpenAI says GPT-6 is better at deciding when to search the web and at finding supporting information.
Keep in mind that these are OpenAI’s own evaluations, not independent benchmarks. The company did not publish public benchmark scores in its announcement.
What OpenAI Says About Safety
OpenAI says GPT-6 is harder to trick into ignoring its safety rules, especially through “multi-turn” attacks, where someone tries to wear the model down over a long conversation. The company says it updated the model’s training for high-risk misuse involving cyberattacks, biological threats and violence, and that the model uses the history of a conversation to spot risks a single message might not reveal.
At the same time, OpenAI says it tried to cut down on unnecessary refusals of harmless requests. It also claims the model is better at admitting its limits, such as saying when it does not have the information or tools to answer properly. Full results are in a technical document called a system card.
What We Do Not Know Yet
The announcement leaves several practical questions unanswered:
- Usage limits. OpenAI did not say what message limits apply to GPT-6 on the free plan or on paid plans.
- Pricing. Subscription prices were not part of the announcement.
- Developer access. There was no word on availability in OpenAI’s API for apps and businesses.
- Accuracy of the visuals. A chart or calculator can look authoritative even when the numbers behind it are wrong. Nothing in the announcement addresses how often the interactive elements contain mistakes.
A Note for Parents
The rollout comes the same week that Common Sense Media’s Youth AI Safety Institute said ChatGPT’s teen version poses an “unacceptable” risk, according to Axios. Its testers ran more than 4,000 prompts on accounts registered to 13- to 17-year-olds.
The group found that ChatGPT generally refused to give harmful instructions, but it missed more than 1 in 4 situations where testers judged a teen should have been pointed to outside help. Testers also said parental alerts did not trigger on newly linked accounts during long conversations about sensitive topics, and that the Study Mode limits were easy to get around.
OpenAI said it welcomes independent testing but that the results do not reflect how its teen safeguards work in practice. It said many of the alert tests ran before parent and teen accounts had finished linking, which can take several hours, and that letting teens leave Study Mode is a deliberate design choice. That testing was about ChatGPT’s teen protections in general, not GPT-6 specifically, but if your kids use ChatGPT, it is worth a conversation about how they use it. Tom Siegel, who heads the institute, advised parents to talk with teens about AI risks and to make sure they have other places to turn for support, saying families cannot rely on a product that is “very early on and still quite buggy.”
How to Get the Most Out of It
- Ask for a format. If you want a table, a calculator or a plain list, say so. The model chooses a format on its own, but you can steer it.
- Check the math. If ChatGPT builds you a savings or loan calculator, test it with numbers you already know before trusting it with real decisions.
- Verify facts. Faster answers are not the same as correct answers. For anything important, follow the links ChatGPT cites or check another source. Our explainer on why chatbots make things up has tips.
- Turn the visuals down if they get in the way. Plain text is still available.
OpenAI frames Intelligent UI as a first step toward software that reshapes itself around what you are trying to do. “Instead of people adapting to software, software will adapt to people,” the company wrote. Whether that turns out to be genuinely useful or just flashy will become clearer once hundreds of millions of people start poking at it this week.
More on Contoh
- Why AI Chatbots Make Things Up — and How to Catch Them When They Do
- ChatGPT Will Start Showing Picture Ads While It Makes Your Images
- AI Chatbots Give a Narrower Range of Information Than a Google Search, Study Finds