ChatGPT Free Gets GPT-6 Today, and Now Your Chatbot Builds You Bill Splitters

SAN FRANCISCO — ChatGPT Free gets GPT-6 today. OpenAI has started rolling out its newest model, GPT-6, to ChatGPT Free and Go users globally, according to Gadgets360's Oct. 8 report, powered by a lower-cost model called GPT-6 Luna. Paid subscribers — Plus, Pro, Business and Enterprise — got the same upgrade a day earlier, on Oct. 7, running a different model: GPT-6 Sol. The headline feature is something OpenAI calls Intelligent UI, and it changes what a ChatGPT answer looks like.

Until now, ChatGPT mostly answered in text, sometimes with a table or a bit of code. Intelligent UI lets it answer with interactive visuals: charts, forms, buttons, diagrams, calculators, bill splitters, even simple games built right into the conversation. The twist is that the chatbot picks the format on its own and builds the answer itself — no special prompt needed — and still uses plain text when that's the better answer. OpenAI's own examples include a recipe shown with its cooking schedule, road-trip stops marked on a map, and letting someone explore a topic by tweaking different inputs. And those interactive elements can start appearing while ChatGPT is still putting together the rest of the answer.

That last part is the sleeper feature. GPT-6 can begin answering while it's still thinking or using tools, then add more findings without another prompt — so you can start reading the response while the model keeps working the rest. In OpenAI's tests, GPT-6 Instant started responding to web-search queries 44 percent sooner than GPT-5.6 Instant, per the company. The model is also better at handling attempts to bypass its safeguards across multi-turn conversations, with added protections against high-risk misuse involving cyberattacks, biological threats and violence, OpenAI says.

There are a few catches worth knowing before you go looking for it. First, the rollout is staged: OpenAI describes it as global but doesn't say when a given account will see it, and its help page notes that if GPT-6 doesn't appear in your model menu, your account may not have it yet. Second, it covers only the Chat tab — OpenAI said it isn't changing the models used for Work or Codex. Enterprise customers may need their workspace administrator to enable the update.

Free users have actually had a taste of Luna already: two weeks ago, OpenAI put GPT-6 Luna inside the ChatGPT desktop app for the free tier, per Notebookcheck. Today's move brings it to the regular Chat tab — the place most people actually use — which is why the staged rollout matters more than the model itself. The people who live in the desktop app were the beta test; everyone else is the launch.

Third, the device details. Notebookcheck reports that older ChatGPT desktop apps for Windows and macOS don't support Intelligent UI at all — OpenAI recommends using the web version instead. On phones, you'll need the latest app version to see the new presentation layer. There is no separate usage quota for Intelligent UI; your plan's usual limits apply. And if the visuals aren't your thing, there's a setting for that: on the web, open Settings and turn off Layout and visuals. But don't expect a full retreat — OpenAI says the setting doesn't switch ChatGPT to a text-only mode, and some visual elements may still show up.

Zoom out and today is the second GPT-6 wave in a month. GPT-6 Astra — the most capable model in the family — arrived in early September for Pro, Business and Enterprise users, and the Luna launch for free users completes the lineup. The timing is also unmistakably competitive: Anthropic shipped Claude Sonnet 5.5 on Sept. 28, a mid-tier model that was 30% faster and up to 30% cheaper per task, and OpenAI answered with its own model-family shake-up. Anthropic, meanwhile, is pushing its own presentation-layer upgrades, adding live dashboards and editable animations to Claude, as we covered last week.

OpenAI isn't done charging for speed, either. Also today, the company launched GPT-6.1 Sol in its Ultrafast service tier — $12 per million input tokens and $60 per million output tokens in the API, six times the model's standard rates, with up to 6x faster token generation in the API and 8x in Codex, per RuntimeWire's Oct. 8 report. Inference latency is becoming a premium line item: you pay the base rate for the model's brain, and a multiplier for how fast it talks.

That's the real stakes of today's rollout. The free tier is where the vast majority of ChatGPT's users live, and Intelligent UI quietly changes the definition of a "chat" with AI — from reading paragraphs to clicking, toggling and playing with the answer. The question hanging over it is the money: OpenAI is extending its most ambitious model to hundreds of millions of free users while shelving models over failed safety tests and losing the people in charge of them. Free, fast and interactive is a great product. Whether the economics behind it hold is the story to watch next.