OpenAI brings GPT-6 to every ChatGPT plan, with clickable answers

The model number is the smaller half of this launch.

Starting this week, a ChatGPT answer can come back as a chart, a diagram, a form or a row of buttons, and every plan gets GPT-6 to draw it.

OpenAI said on October 7 that it is launching GPT-6 to everyone in ChatGPT, on free and paid plans globally, in its deployment safety notes. The two models, GPT-6 Sol and GPT-6 Luna, replace GPT-5.6 Sol and GPT-5.6 Luna in ChatGPT.

Paid plans got it Wednesday, free plans start today

According to The Verge and 9to5Mac, Sol and the new Intelligent UI reached Plus, Pro, Business and Enterprise on Wednesday. Luna, which The Verge calls the more efficient model, begins rolling out to Free and Go on Thursday, October 8.

OpenAI says GPT-6 Sol and Luna in Codex and via ChatGPT Work are still running previously released versions.

Intelligent UI is the part you will notice. The Verge reports that answers now mix text with diagrams, charts, forms, tappable buttons and more, and that OpenAI trained GPT-6 to judge when a visual beats a paragraph. TechCrunch describes examples such as a diagram of how a wing creates lift and a map for a multi-day hike. 9to5Mac reports ChatGPT can also begin answering while it is still thinking.

Notebookcheck reports there is no separate quota for the visual answers. Older Windows and macOS desktop apps don't support them, and the web setting called "Layout and visuals" tones them down without switching them off.

The answer is turning into the app

Here's the thing. A chatbot that builds a calculator on request competes with every small web tool you used to search for.

The Verge says you can ask ChatGPT to build tools inline, such as a retirement savings calculator or a bill splitter. If you publish calculators, explainers or simple utilities, that traffic now has a rival inside the chat window. If you build products on top of ChatGPT, your users will start expecting answers they can click.

Yes, it is another model number. The interface is what changes for everyone at once.

Read the safety notes before you roll it out

OpenAI treats both models as High capability in cybersecurity and in biological and chemical risk, according to its safety page.

The same page lists regressions against the GPT-5.6 models. Sol shows a statistically significant regression on self-harm evaluations. Luna regresses on self-harm, gore and sexual content. OpenAI says it reviewed the failures and judged them "borderline but still generally safe."

On its under-18 evaluations, OpenAI reports regressions for both models on age-restricted content, sexual content and emotional reliance. It says an extra classifier-based block for teens is not reflected in those results.

Some numbers moved the right way. On a test of whether a model works around a restriction it was warned about, where lower is better, OpenAI reports:

Luna fell from 27% to 15.9%.

Sol fell from 34% to 28%.

If you manage ChatGPT for a team, those lines belong in your rollout note. The teen results land on the same day as Common Sense Media's rating of ChatGPT for Teens.

Watch the developer side next

None of the OpenAI pages we opened gives API availability or pricing for GPT-6 Sol or Luna. That is the next number to watch if you build on OpenAI, because a cheap Luna in the API would reset what small models cost. Anthropic already priced Claude Haiku 5.5 against it.

Then watch your own analytics. If fewer people click through to your tools this month, you will know where they went.

Open ChatGPT on the web today and ask for something you would normally build a spreadsheet for.