GPT-6 is OpenAI's current model generation, and it is only now reaching most ChatGPT users. OpenAI introduced GPT-6 Astra in September 2026, followed by the faster, cheaper GPT-6 Sol and Luna on September 22. On October 7 it began rolling GPT-6 out to ChatGPT's Chat tab alongside a new feature called Intelligent UI. This guide explains what OpenAI has said is new, which model you are likely using, and how to write prompts that make use of it. Everything described as OpenAI's claim comes from its own announcements, linked at the end. The prompting advice is our own practical guidance, not something OpenAI has published.

What Is GPT-6?

GPT-6 is a family of models rather than a single one. GPT-6 Astra is the top of the family, which OpenAI describes as its most intelligent and aligned model and recommends for the most demanding projects. GPT-6 Sol and GPT-6 Luna were trained with similar methods but are faster and more affordable, offering different balances of capability and cost. Astra, Sol and Luna are public model names, not internal codenames.

The family has already moved on once: OpenAI announced GPT-6.1 Sol on September 29, 2026. This guide covers the GPT-6 announcements above and does not go into 6.1, so check OpenAI's own page for what changed there.

Which GPT-6 Model Are You Using?

In ChatGPT's Chat tab, OpenAI says GPT-6 is powered by Sol for Plus, Pro, Business and Enterprise plans, and by Luna for Free and Go plans. The rollout began for paid plans on October 7, 2026 and expands to Free and Go from October 8. Enterprise availability depends on workplace admin settings. If you do not see it yet, the rollout is gradual, so check again later.

This update applies to the Chat experience only. OpenAI states that the models powering ChatGPT's Work mode and Codex are not changing as part of this release. Developers can call the models in the API as gpt-6-sol and gpt-6-luna.

At launch OpenAI listed API pricing per million tokens of $2 input and $10 output for Sol (down from $4 and $20 for GPT-5.6 Sol) and $0.10 input and $0.50 output for Luna (down from $0.20 and $1.20 for GPT-5.6 Luna). Prices can change, so confirm on OpenAI's pricing page before budgeting.

Check Which Model You Have The model behind a conversation depends on your plan and on rollout timing, so two people asking the same question may be talking to different models. If a result differs from what this guide describes, that is a likely reason.

What OpenAI Says Is New

These points summarise OpenAI's own announcements. Benchmark figures in those posts come from OpenAI's evaluations, so treat them as indicative rather than independent.

Fewer factual mistakes. On OpenAI's internal factuality evaluation, built from real conversations where users had flagged errors, GPT-6 Sol makes about half as many mistakes as its predecessor. OpenAI notes those conversations are not representative of typical use, where errors are rarer. Fewer is not none, so anything consequential still needs checking.

More honest about its limits. OpenAI says GPT-6 is better at recognising and communicating what it cannot do, including when it lacks the information or tools needed to answer.

A different communication style. OpenAI says Sol and Luna aim for more clarity, less jargon, fewer odd turns of phrase and slightly shorter answers overall without losing substance.

Answering while thinking. ChatGPT can now begin answering while it is still working on the problem, building the answer across several partial responses. For questions that need web search, OpenAI reports GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant, and makes better decisions about when to look something up.

Coding, professional work and computer use. OpenAI reports gains on its published benchmarks, at lower cost per task than the previous generation. The results are vendor-run and we have not reproduced them.

Intelligent UI: Prompts for Interactive Answers

Intelligent UI is the most visible change in ChatGPT. Instead of always replying in text, GPT-6 can compose responses from text, graphics, tappable buttons, forms, charts and interactive elements, choosing the format based on your question. OpenAI's own examples include an interactive breakdown of a 7-speed bike, a Sunday roast plan that updates when you change the guest count, a road trip shown on a map, explainers for the central limit theorem and the Monty Hall problem, and small tools such as a savings calculator, a bill splitter and a retro game. OpenAI also says that when plain text is the most useful answer, ChatGPT can simply give you that.

The prompts below are starting points modelled on those examples. We have not tested every variation, and OpenAI says design judgment is still being improved, so expect uneven results.

Break down how [a mechanism, e.g. a heat pump] works as an interactive explainer I can click through one component at a time. Keep each explanation to two sentences.

Build a calculator where I can adjust [inputs, e.g. monthly saving and years] and see how [outcome, e.g. the final balance] changes. Use realistic defaults and show your assumptions clearly.

I am planning [dinner, trip or event] for a number of people I have not settled on yet. Make the plan adjustable so quantities and timings update when I change the headcount.

Explain [a statistical or scientific idea] with something I can interact with, so I can change one input and see what happens.

Make a simple bill splitter for [number] people with different amounts per person and an optional tip.

Compare [option A] and [option B] side by side on the factors that matter for [my situation], then give your recommendation in plain text underneath.

Make a quick game I can play in this chat to practise [skill, e.g. times tables or Spanish vocabulary]. Keep score and get gradually harder.

Answer this in plain text only, with no interactive elements: [question].

Prompting Tips for GPT-6

None of the announcements we reviewed describe a new prompting style. The tips below follow from the changes OpenAI did describe and from general good practice. They are our suggestions, not tested rules.

Ask for the depth you want. If answers are slightly shorter by default, say so when you need more: "walk me through the reasoning" or "aim for about 800 words" removes the guesswork.

Say what should come first. Because ChatGPT may now answer in stages, tell it the order you want: "lead with the recommendation, then the reasoning, then the risks."

State hard constraints explicitly. Numbers, deadlines, audience and budget are not things the model can safely infer. A model that makes fewer mistakes still cannot know your budget.

Ask what it could not verify. Since OpenAI says GPT-6 is better at flagging its limits, invite it: "tell me which parts of this you are unsure about or could not check." Then check those parts yourself.

Choose the format deliberately. With Intelligent UI you can now ask for a tool, a table or plain text. Naming the format you want avoids getting an interactive widget when you needed something to paste into an email.

General-Purpose Prompt Templates

These templates are not specific to GPT-6. They are general-purpose prompts that work on any recent model and are included so you have a ready set to test.

I need to decide between [option A] and [option B] by [date]. My hard constraints are [constraint 1] and [constraint 2]. Recommend one, explain the trade-offs, and tell me what would change your recommendation.

Review this plan: [paste]. Identify the one assumption it depends on most, explain what happens if it is wrong, and suggest a cheap way to test it before I commit.

I need to tell [audience] that [difficult news]. Draft the message, then list the three questions they are most likely to ask and give a short answer to each.

Review this code: [paste]. Tell me whether the design will cope with [future requirement], what would break first, and the smallest change worth making now.

I am debugging [describe the problem]. Do not guess a fix. Explain what you would check first, what each result would tell us, and how to narrow down the root cause.

I am writing [describe the piece]. Before drafting, ask me the questions you would need answered to do this well. Only ask questions that would change what you write.

Coming From GPT-5

If you built a prompt library for GPT-5, there is no need to rewrite it wholesale. Re-run your five most-used prompts on GPT-6, compare the results with what you got before, and change a prompt only where you see a real difference. Keep the rest as it is. Our GPT-5 prompts guide still covers the principles those prompts were built on.

OpenAI also published a developer-focused post, A practical guide to building with GPT-6, on October 2, 2026. We have not reviewed it here, but it is the place to look if you are building on the API.

Sources and Verification

Facts about the GPT-6 family, availability, pricing and Intelligent UI in this guide come from two OpenAI posts, checked on October 8, 2026: Introducing GPT-6 Sol and Luna and GPT-6 and Intelligent UI for everyone. Model lineups, plan availability and prices change quickly, so check those pages and OpenAI's release notes for the latest.

Common Mistake Treating a launch announcement as independent testing. Benchmarks and cost comparisons in model launch posts are run by the vendor. Test the model on your own tasks before changing a workflow around it.

For the broader principles behind these prompts, see our prompt engineering guide, and for the previous generation, our GPT-5 prompts guide.

Frequently Asked Questions

What is GPT-6?

GPT-6 is OpenAI's current model family, made up of GPT-6 Astra, GPT-6 Sol and GPT-6 Luna. Astra is OpenAI's top model, while Sol and Luna are faster, more affordable models trained with similar methods. OpenAI began rolling GPT-6 out to ChatGPT's Chat tab on October 7, 2026.

Which GPT-6 model does ChatGPT use?

According to OpenAI, GPT-6 in ChatGPT's Chat tab is powered by GPT-6 Sol on Plus, Pro, Business and Enterprise plans, and by GPT-6 Luna on Free and Go plans. The models behind ChatGPT's Work mode and Codex were not changed in that release.

Is GPT-6 available on the free plan?

OpenAI said GPT-6 would start rolling out to Free and Go users on October 8, 2026, powered by GPT-6 Luna. The rollout is gradual, so it may not appear in every account immediately.

What is Intelligent UI in ChatGPT?

Intelligent UI is a GPT-6 feature that lets ChatGPT answer with interactive elements such as graphics, buttons, forms, charts and small tools, as well as plain text. ChatGPT chooses the format based on your question, and you can ask for plain text if you prefer it.

Do I need to rewrite my GPT-5 prompts for GPT-6?

Probably not wholesale. Re-run your most-used prompts on GPT-6, compare the results, and adjust only where you see a real difference. The announcements we reviewed do not describe a new prompting style.

How much does the GPT-6 API cost?

At launch OpenAI listed GPT-6 Sol at $2 per million input tokens and $10 per million output tokens, and GPT-6 Luna at $0.10 and $0.50 respectively. Prices can change, so check OpenAI's pricing page for current figures.