Why ChatGPT Now Draws Its Answers

OpenAI began putting GPT-6 into the ChatGPT app on 7 October, and the free and Go tiers get it on 8 October. The part most people will notice is not a smarter answer. It is that the answer can now arrive as a chart, a form, a diagram or a small tool instead of a wall of text. OpenAI says the model starts answering web search questions 44% sooner on average than the version before it, because it answers while it is still thinking. Singapore sends more ChatGPT messages per person than any other country in OpenAI's own data, so this arrives here early. My own view is that the shape of an answer is a bigger deal than it sounds, and that a tidy answer is also easier to trust than it deserves.

OpenAI's announcement on 7 October moved GPT-6 into the ordinary Chat tab.

Paid plans, which are Plus, Pro, Business and Enterprise, get it first, running on a model called GPT-6 Sol.

Free and Go users get a model called GPT-6 Luna from 8 October.

OpenAI says the change reaches more than 1.2 billion people who use ChatGPT every week.

The feature is called Intelligent UI.

It lets the model decide how an answer should be laid out, based on what you asked.

Ask for a comparison and the two things can sit side by side.

Ask how the gears on a bicycle work and the demonstration shows a diagram with buttons for the frame, the wheels and the brakes.

Ask to plan a drive and the stops can appear on a map, with a note on where to stop.

Ask about cooking a roast and the timings can sit next to the recipe.

You can also ask for something you use rather than read.

OpenAI's own examples are a calculator to see how savings might grow, a bill splitter for dinner with friends, and a small game.

None of that is a new kind of intelligence.

It is a change in what an answer looks like.

The second change is about waiting.

With the reasoning models that came before, you asked, then waited while the model thought, then got the whole answer at once.

GPT-6 was trained to start answering while it is still working, and to add more detail as it goes.

OpenAI says that for questions which need a web search, GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant.

It also says that in one of its internal tests, a setting called GPT-6 Extra High began answering in about the same time as the older mid-level setting while scoring better than the older top setting.

That last number is OpenAI marking its own homework, so I would read it that way.

The 44% is easier to feel.

If you have ever watched a chat window sit still for ten seconds, you know why.

There is a local reason to pay attention.

OpenAI publishes country data on how ChatGPT is used.

For the second quarter of 2026, covering April to June, it ranked countries by the number of ChatGPT messages sent per person.

Singapore came first out of 147 countries.

The United States sits at 51st, Malaysia at 59th, Indonesia at 101st and India at 109th.

The Singapore AI Observatory translated the release and added its own note pointing out that this is the third set of independent data to place Singapore at or near the top.

It cites Anthropic putting Singapore first among 116 regions in March, and Microsoft putting it second worldwide in January.

For anyone running something here, that has a plain meaning.

Getting more people to open a chat window is no longer the problem.

What they do once it is open is.

The same note carries a limit worth repeating: this data covers personal accounts, not company ones, so it says nothing about how much of the work inside a business is being handed over.

If you run something small, the useful part is not the polish.

It is that you can ask for the tool you need instead of a description of it.

A shop owner comparing two supplier prices can ask for a table that changes when she edits a number, instead of reading an answer and then redoing the sums by hand.

Someone arranging leave for a small team can ask for something that shows where the clashes are.

Someone collecting quotes can ask for a form that adds the numbers up as they come in.

The question shifts from what does the answer say to what can I change myself to test it.

That is the part worth learning this week, and it takes an afternoon, not a course.

Two things hold me back.

First, the announcement does not say how sources will appear when the answer is a chart or a tool.

Search Engine Journal raised that point plainly, and it is the right one: with a calculator or a comparison on the screen, the question of where the facts came from is left open.

A neat chart is easier to believe than a paragraph.

That is the risk, not the convenience.

Second, the rollout is uneven.

The top thinking setting, called Astra, does not support the new interface yet, and the older desktop apps for Mac and Windows do not either.

On the web you can switch the visuals off in the settings, and OpenAI's help page warns that some elements can still show up anyway.

So two people on the same paid plan may see different things, and it is worth knowing that before you blame your own screen.

I think the interesting move here has little to do with intelligence.

For two years the story was that the model got better at answering.

This release is about the shape of the answer.

For decades software was built around a task, and we learned where the buttons were.

OpenAI is now saying the screen should be built around your question instead.

That is a real change, and it is early.

My worry is that a nicer answer feels more finished than it is.

A paragraph leaves room for doubt.

A clean chart with one number on it does not, and the number can still be wrong.

I would use the new tools to check my own thinking, not to skip it.

Build the calculator, then check the arithmetic.

Ask for the comparison, then ask where the prices came from.

Access was never the advantage in Singapore, because almost everybody here has it.

Knowing what to ask, and knowing what to check, still is.

So the question I keep coming back to is whether a screen that explains itself helps me understand more, or quietly means I ask less.