Atlas AI Resources

← Back to AI News

AI products

GPT-6 Starts Reaching Free ChatGPT Users. Answers Now Come With Charts and Buttons.

Published October 8, 2026

On October 7, OpenAI started switching ChatGPT’s everyday chat to GPT-6, beginning with paying subscribers. Starting today, October 8, the rollout expands to the free and low-cost Go plans. The headline feature, which OpenAI calls Intelligent UI, lets ChatGPT answer with diagrams, charts, forms, tappable buttons, and small working tools instead of only text. OpenAI’s own safety report says the new models improved in several areas and slipped in a few.

What happened

GPT-6 replaces GPT-5.6 in ChatGPT’s Chat tab. Plus, Pro, Business, and Enterprise subscribers get a version called GPT-6 Sol. Free and Go users get GPT-6 Luna, which The Verge describes as the more efficient version. OpenAI says the models behind ChatGPT Work and Codex, its coding tool, are not changing in this release.

“UI” is short for user interface: the panels, buttons, and charts you see and tap. OpenAI says it trained GPT-6 to decide, question by question, whether text, a visual, an interactive piece, or a mix works best. The pieces come from a library of ready-made building blocks, and a compiler (software that turns the model’s output into those on-screen parts) draws them as the answer is being written, so you are not staring at a blank screen.

OpenAI’s examples include a tappable diagram of a seven-speed bicycle, roast timings laid out next to a recipe, road-trip stops on a map, and tools you can ask for, like a savings calculator, a bill splitter, or a simple game. TechCrunch reports that OpenAI product manager Aarush Selvan demonstrated diagrams explaining how an airplane wing creates lift, and that users will be able to dial the visuals back.

GPT-6 can also start answering while it is still thinking. OpenAI says that for questions that need a web search, GPT-6 Instant begins answering 44% sooner on average than GPT-5.6 Instant. That figure comes from OpenAI’s internal testing.

Why it matters to a normal person

OpenAI says more than 1.2 billion people use ChatGPT every week, so a change in how answers look reaches a lot of people at once. A diagram you can tap, or a calculator you can adjust, can make a hard idea click faster than a wall of text.

The safety details matter too. OpenAI’s system card (the safety report it publishes with a model) says GPT-6 better resists jailbreaks, meaning tricks that push a model to ignore its safety rules, makes fewer factual errors on its hard test sets, and refuses harmless requests less often. It also reports statistically significant slips. GPT-6 Luna, the free model, scored worse on self-harm, gore, and sexual-content tests, and both versions slipped on some tests for teen users. OpenAI says it reviewed those failures, judged them generally low severity, and adds an extra filter for users it believes are under 18 that those scores do not include.

What it does not mean

A polished interface is not the same as a correct one. A savings calculator built on the spot can look like a finished app and still use a wrong formula or an assumption you never chose. The model underneath can still make mistakes, so check numbers that matter, just as you would with a text answer.

It does not mean free and paid users get the same answers. Free users get Luna, not Sol.

And the speed and quality claims are OpenAI’s own. Independent testing has not been published yet, and OpenAI itself says there is “still work ahead to improve the model’s design judgment.” OpenAI describes the free-plan launch as a rollout that starts today, so it may take time to reach every account.

Sources