GPT-6 is free now, with an asterisk. Since October 8, OpenAI has been rolling it out to ChatGPT Free and Go users, but the model doing the work on those plans is GPT-6 Luna, the most efficient member of the family. Pay for Plus, Pro, Business or Enterprise and the same Chat tab runs on GPT-6 Sol instead. GPT-6 Astra, OpenAI's top model, isn't available on Free or Go at all. If you mostly live in Codex or ChatGPT Work, you can ignore the whole thing: OpenAI says the models powering Work and Codex "are not changing as part of this release."

The rollout came in two steps. According to OpenAI's announcement and its October 7 ChatGPT release notes, GPT-6 with Intelligent UI reached Plus, Pro, Business and Enterprise first, on October 7, and expanded to Free and Go "starting tomorrow," on October 8. Those are OpenAI's dates; its developer community post went up at 03:13 on October 8, Beijing time. OpenAI calls the rollout global but won't say when any particular account gets it. The pricing page has already caught up. It promises "Unlimited text chats with GPT-6" on the $0 Free plan, and it shows GPT-6 Luna as included for Free and Go, with GPT-6 Sol, GPT-6.1 Sol and GPT-6 Astra marked as not included.

The help center is lagging behind. OpenAI's article on ChatGPT models still says Free and Go use GPT-5.6 Luna for Instant and that GPT-5.6 Luna "also powers Think." That contradicts the announcement, the release notes, the developer community post and the system card, which all say GPT-6 Luna, and we're going with the announcement. Accounts are being switched in batches, though, so yours may simply not be on GPT-6 Luna yet. Think is the murkier part. OpenAI hasn't said whether Think on Free and Go moves to GPT-6 Luna, and its Go help page still reads "Think uses GPT-5.6 Luna."

The plan-by-plan split

Here's how OpenAI divides up the Chat tab:

  • Free ($0) and Go ($8/month): GPT-6 Luna, with Think for harder questions. No GPT-6 Sol and no Astra.
  • Plus ($20/month): GPT-6 Sol at Instant, Medium and High, without Extra High or Pro.
  • Pro (from $100/month): GPT-6 Sol all the way from Instant to Extra High, plus GPT-6 Pro, which runs on GPT-6 Astra.
  • Business and Enterprise: the same thinking levels as Pro, GPT-6 Pro included. On Enterprise, access depends on admin settings.

Don't bother looking for the word "Sol" in your menu. Paid users see "GPT-6," which, according to OpenAI's help article, replaces "Latest." If Latest was your pick, you've been moved to GPT-6 at the same effort. If you had deliberately chosen GPT-5.6 Sol, you stay on GPT-5.6, and nothing you can currently select is being retired.

Step outside Chat and the lineup changes again. ChatGPT Work and Codex let you pick GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna, models OpenAI says "are not available in regular ChatGPT conversations." Plus subscribers get GPT-6 Astra in Work and Codex, but not GPT-6 Pro in Chat. Free and Go get limited Work and Codex access in the desktop app, and OpenAI's Codex models page points those users to GPT-6 Luna.

Three Sols, two Lunas and no Terra

Most of the confusion comes from OpenAI's naming, and it's easier to follow by date. GPT-6 Astra came first, on September 3. It's OpenAI's most capable model, and in Chat it shows up as GPT-6 Pro on Pro, Business and Enterprise; Plus gets it only in Work and Codex, and Free and Go don't get it. On September 22, GPT-6 Sol, the middle model, and GPT-6 Luna, the most efficient one, arrived in Work, Codex and the API, and Free and Go users could try Luna in the desktop app. A week later, on September 29, OpenAI shipped GPT-6.1 Sol for Work, Codex and the API. OpenAI says it "nearly matches GPT-6 Astra's intelligence on agentic coding, computer use, and professional work," and it is "not yet available in Chat."

What landed in Chat this week is a newer cut of Sol and Luna. OpenAI's system card for the release, published October 7 on its Deployment Safety Hub, says the new models "will replace GPT-5.6 Sol and GPT-5.6 Luna in ChatGPT," while people using GPT-6 Sol and GPT-6 Luna in Codex and ChatGPT Work "are still using previously released versions." OpenAI tells them apart by month: October for Chat, September for Work and Codex.

So there are now three Sols. GPT-6 Sol (October) powers GPT-6 in Chat on paid plans, and you can't select it in Codex. GPT-6 Sol (September) is the one in the Work and Codex pickers, and it's what the API sells as gpt-6-sol. GPT-6.1 Sol is newer and stronger, lives only in Work, Codex and the API as gpt-6.1-sol, and is the model our GPT-5.5 retirement guide recommends for Codex.

As for Terra, there is no GPT-6 Terra. The Terra that keeps turning up in searches is GPT-5.6 Terra, an older GPT-5.6 model that OpenAI pitches as a balance of capability, speed and cost. You can't pick it in standard ChatGPT conversations. Paid plans can use it in Work and Codex, Free and Go get limited access to it in Work and Codex on desktop, and it's on the API as well.

Where Sol pulls ahead of Luna

OpenAI describes both Chat models as "tuned for everyday conversation," and both get Intelligent UI, so an answer can include charts, buttons, forms or a small tool like a calculator. On a simple question, a Free user and a Plus user may not notice much difference. The gap opens when you want the model to think harder. Paid plans can push GPT-6 Sol to Medium and High, and Pro, Business and Enterprise can go to Extra High, while Free and Go get Instant plus Think.

How much the model can hold in view also tracks what you pay. OpenAI's pricing page lists a total context window for Instant of 27K on Free, 54K on Go and Plus, and 128K on Pro. For reasoning, it's 256K on Go and Plus, 400K on Pro, and "Varies" on Free.

OpenAI's own numbers show a real quality gap between the two models. In the October system card's length-adjusted HealthBench Professional results, GPT-6 Sol (October) scores 55.2 against 48.2 for GPT-6 Luna (October). Both still beat their GPT-5.6 predecessors, and OpenAI says both make fewer factual errors than the GPT-5.6 models on nearly all of its hallucination metrics. On speed, OpenAI says GPT-6 Instant starts answering questions that need a web search 44% sooner on average than GPT-5.6 Instant. That's a measure of when the answer begins, not when it ends, and OpenAI doesn't split it out for Sol and Luna.

Intelligent UI runs from Instant through Extra High but not on the Pro option, which still uses GPT-6 Astra, and it has no quota of its own. If you don't see it, your app is the likely culprit: OpenAI says the older ChatGPT desktop apps for macOS and Windows don't support it and recommends the web. If the visuals get in your way, you can switch off Layout and visuals in Settings on the web, although some visual elements may still appear.

Generous for chatting, vague about thinking

OpenAI is much more open about Codex limits than about Chat. Free and Go get "unlimited everyday text chats," subject to abuse-prevention safeguards, with separate limits on file uploads, image generation, voice, data analysis and other tools. Go's limits are higher, but OpenAI doesn't publish the numbers. On Plus, Pro, Business and Enterprise, OpenAI says only that Thinking limits depend on your plan, with no figure for GPT-6 Sol. If you hit one, ChatGPT may carry on with another available model.

GPT-6 Pro is capped even on Pro. On Pro $200, running out of the weekly GPT-6 Pro limit drops you automatically to GPT-5.6 Thinking at Medium. On Pro $100, GPT-6 Pro and GPT-5.6 Sol Pro draw from one shared weekly allowance. OpenAI doesn't give message counts for either.

Work and Codex share a separate allowance, and this is where OpenAI does offer estimates. For Plus and Business Standard seats, it puts local messages per five-hour window at 5–45 for GPT-6 Astra, 15–160 for GPT-6.1 Sol, 15–150 for GPT-6 Sol and 350–3,000 for GPT-6 Luna. They're estimates rather than guarantees, and weekly limits may also apply. Pro 100, Pro 200 and Pro 500 currently have no five-hour limit in Work and Codex.

API prices, and the version question

For developers, these are per 1 million tokens at standard processing for context under 272K, taken from OpenAI's API pricing page and model pages and checked on October 9, 2026:

  • gpt-6-astra: $10 input, $1 cached input, $50 output.
  • gpt-6.1-sol: $2 input, $0.10 cached input, $10 output.
  • gpt-6-luna: $0.10 input, $0.01 cached input, $0.50 output.
  • gpt-6-sol: $2 input and $10 output, according to its model page, which now points to gpt-6.1-sol as "the newer Sol model."

Batch and Flex cost half the standard rate and Fast mode costs double. All of this matches the GPT entry in our Model directory.

What OpenAI hasn't said is whether the API's gpt-6-sol and gpt-6-luna serve the September or the October versions. The pricing page also lists chat-latest at $5 input and $30 output, and the docs say it "points to the latest Instant model currently used in ChatGPT," without saying which GPT-6 model that is right now. Until OpenAI clears that up, we wouldn't treat gpt-6-sol or chat-latest as an exact copy of the GPT-6 in your ChatGPT plan, and we'd test either one before building on it.

If you mostly use Codex

According to OpenAI, nothing changes for you. Codex and Work keep the September versions of GPT-6 Sol and GPT-6 Luna, along with GPT-6.1 Sol and GPT-6 Astra where your plan includes them, so there's no reason to touch your settings for GPT-6. The date that matters is still October 14, when GPT-5.5 leaves ChatGPT, ChatGPT Work and Codex, and any config that pins gpt-5.5 should be replaced before then. OpenAI's Codex models page recommends GPT-6.1 Sol for complex work and Luna for focused, repeatable tasks, and if your plan has Astra, that's the one to hand the hardest jobs. Our guide "GPT-5.5 Leaves Codex on October 14: What to Switch to and What to Fix" walks through the config changes, and "From $8 to $500: Every Major AI Coding Plan Compared (Oct 2026)" covers what each plan costs.

When upgrading makes sense

If you only chat casually on Free, there's no reason to pay just to get GPT-6. Once the rollout reaches your account, GPT-6 Luna is yours at no cost. Go doesn't change the model either: for $8 you stay on Luna and get higher tool limits, which mainly helps if you keep running into them. Plus at $20 is the first plan that moves you to GPT-6 Sol, with Medium and High thinking and limited GPT-6 Astra in Work and Codex. If you want the strongest answers inside Chat itself, only Pro, Business and Enterprise get GPT-6 Pro (Astra), and even there it has its own limits.

If you're on a paid plan, "GPT-6" should already be in your model menu. If it isn't, OpenAI says the rollout may not have reached your account yet.