GPT-6.1 Sol Explained: Your Plus Plan Includes It, Your ChatGPT Does Not

OpenAI released GPT-6.1 Sol on September 29, 2026, and exactly a week after GPT-6 Sol it put another new model somewhere most paying customers will never look. If you pay for ChatGPT Plus, your plan now includes GPT-6.1 Sol. If you open ChatGPT and go looking for it, it is not there. OpenAI's own help centre states the reason in one sentence, and the pricing page you would check first does not mention it.

Key Takeaways

• It is not in chat, and OpenAI says so plainly. Its ChatGPT Work and Codex help article: "GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna are models for ChatGPT Work and Codex. They are not available in regular ChatGPT conversations."
• Plus is the cheapest way in, at $20 a month. OpenAI's docs: the rollout "includes Plus, Pro, Business, Enterprise, and Edu," and "Free and Go are not included at launch."
• Free and Go get nothing this time. That is a change from GPT-6 Luna, which Free and Go could reach in the desktop app. GPT-6.1 Sol closes that route.
• No price moved. Free is still $0, Go $8, Plus $20, Pro from $100. What you get is an allowance, not a purchase.
• On Plus the allowance is 15 to 160 messages per five-hour period in Work and Codex, against 5 to 45 for GPT-6 Astra and 350 to 3,000 for GPT-6 Luna. Pro plans "currently have no five-hour limit."
• The pricing page looks like it disagrees, and it does not. Its GPT-6.1 Sol row reads Free No, Go No, Plus Yes, Pro Expanded and never says which part of the app, even though the same table carries "Limited access in Work and Codex on desktop" on another row.
• The selling point is price, not capability. OpenAI says GPT-6.1 Sol "nearly matches GPT-6 Astra's intelligence on agentic coding, computer use, and professional work at one-fifth of Astra's standard input and output token prices."
• The fast version is not here yet. OpenAI: "Ultrafast support for GPT-6.1 Sol is coming later," and in the announcement, "in the coming days."

Do You Get GPT-6.1 Sol? Availability by Plan

Every consumer plan, read against OpenAI's Codex and ChatGPT Work model docs and its pricing docs on October 2, 2026. The last column is the same answer on every single row, which is the point of the table.

OpenAI attaches two caveats itself. On Enterprise and Edu, "the plan keeps GPT-6.1 Sol off by default until an administrator enables it." And across the board, "availability depends on your plan, workspace settings and rollout access," so a Plus subscriber may not see it on day one either.

Where It Actually Lives

OpenAI publishes a per-surface switch list for each model, and comparing two of them settles the question without interpretation. GPT-6 Astra has rows for the chat app; GPT-6.1 Sol does not have them at all.

So reaching GPT-6.1 Sol means switching out of Chat into Work, which OpenAI describes as "an agent designed for longer, multi-step work and finished deliverables," or into Codex, which is "dedicated to software development and technical work." Neither is the chat box on the ChatGPT home screen.

Why the Pricing Page Is Misleading, and How to Read It

OpenAI's ChatGPT pricing page has a Models section that ticks four GPT-6 models for Plus, and reading it the obvious way will send you hunting through a model picker for something that is not in it. Here is the GPT-6.1 Sol row exactly as that table publishes it, read October 2, 2026:

That row is true. It is a statement about your plan, not about where the model appears. The table proves it is capable of saying otherwise when it wants to: two rows further down, the GPT-5.6 Terra cells for Free and Go read "Limited access in Work and Codex on desktop," spelled out in full. On the GPT-6.1 Sol row, that qualifier is simply absent.

The practical rule for any OpenAI plan table: a tick in a Models row tells you your plan is allowed the model, not that you can find it in the chat window. For GPT-6.1 Sol, four separate OpenAI pages say Work and Codex, not Chat: the help article, the model docs, the pricing docs and the release note. None of them is the page a customer checks first.

How Much Of It Do You Get?

On Plus, OpenAI estimates 15 to 160 local messages per five-hour period. That figure comes from its Codex pricing docs, where the table is headed "local messages per five-hour period for Plus and Standard Business," and OpenAI notes that "cloud tasks may use more of your allowance than local messages" and that the numbers "are not fixed message limits."

Read against the row above it, this is the release's actual consumer benefit: GPT-6.1 Sol gives a Plus subscriber roughly three times as many messages as Astra at the same capability rating OpenAI gives Astra in its own model docs. Pro plans sit outside this table entirely, because "Pro plans currently have no five-hour limit."

In credit terms OpenAI prices GPT-6.1 Sol at 50 credits per million input tokens, 2.5 per million cached input tokens, and 250 per million output tokens. Astra is 250, 25 and 1,250 for the same three. The cached-input number is the one that moved most: it is half GPT-6 Sol's 5 credits.

Is It Actually Better, or Just Cheaper?

Cheaper is the headline, and OpenAI is direct about it: GPT-6.1 Sol "nearly matches GPT-6 Astra's intelligence on agentic coding, computer use, and professional work at one-fifth of Astra's standard input and output token prices." Its own one-line description of the model leads with cost: "Near-Astra performance for complex work at a lower cost than Astra."

Two of its published numbers are worth an everyday reader's attention.

Factual accuracy went up measurably. OpenAI says its "largest factuality improvement over GPT-6 Sol comes at low reasoning effort, where it reduces the share of responses containing a factual error from 11.4% to 7.7% -- a reduction of approximately 32%." The caveat is OpenAI's own and matters: that test uses "de-identified conversations where users flagged an earlier model's error," which OpenAI says "are not representative of typical usage."

Astra is still the top of the range. On the scientific-workflow benchmark OpenAI reports, "GPT-6 Astra still achieves the highest score among the models tested at 68.1%, and should be used for the most difficult scientific research tasks." GPT-6.1 Sol's pitch there is cost: $5.47 per task at maximum effort against $23.80 for Astra.

For what it is for, OpenAI's own non-technical examples are the clearest guide: consider GPT-6.1 Sol "for complex projects where cost matters, such as creating a board presentation from financial results or building a website from a product brief."

The Naming Problem Is Getting Worse

There are now four GPT-6 models and a GPT-5.6 family running alongside them, and the one in your chat window is from the older family. As of October 2, 2026:

"GPT-6.1 Sol" and "GPT-6 Sol" are a week apart and differ by a decimal. "GPT-5.6 Sol" is a different family and is the one you are actually talking to if you pay $20. OpenAI has not retired any of them.

One more date to keep: GPT-5.5 retires on October 14, 2026 "from ChatGPT, ChatGPT Work, and Codex on all plans." Note which model OpenAI names as the replacement, because it is not this one: "On Plus, Pro, Business, Enterprise, and Edu plans, choose GPT-6 Sol," and on Free and Go, GPT-6 Luna in the desktop app. That only affects people who pick GPT-5.5 by hand.

Should You Care?

Care if you pay for Plus or Pro and you produce documents, decks or spreadsheets. This is the best model-per-allowance deal OpenAI has put on a $20 plan, and it is in ChatGPT Work, which you reach from the same app you already pay for. If you have never opened the Work toggle, this release is the reason to.

Skip it if you use ChatGPT for questions, writing help, summaries or search. Nothing on your plan changed. Your chat still runs GPT-5.6, your limits are the same, and there is no setting to turn on. Our guide to whether ChatGPT Plus is worth it does not move because of this release.

Do not upgrade from Free or Go for this. The route Free and Go had into the GPT-6 family, GPT-6 Luna in the desktop app, does not extend to GPT-6.1 Sol. Going from $0 to $8 buys you nothing here; $0 to $20 buys you a model in a tab built for long-form work.

What OpenAI Has Not Published

No date for Chat. The announcement says GPT-6.1 Sol "is not yet available in Chat," and "not yet" is the whole of the guidance. No OpenAI page we read gives a date, a plan or a condition for it arriving there.

No Ultrafast yet. OpenAI's docs say "Standard and Fast modes are available at launch. Ultrafast support for GPT-6.1 Sol is coming later." The announcement puts it "in the coming days" with "up to 8x faster token generation compared to its standard speed in Codex." That was September 29; checked October 2, the docs still say "coming later."

The model release notes page still has no GPT-6 entry of any kind. OpenAI keeps a model release notes page for exactly this class of change. Read through in the newest copy available, captured September 27, 2026, its most recent entry is a Model Spec update dated August 18, 2026. Every GPT-6 model, Astra included, has been announced somewhere else. That is a fact about that page, not about the product.

Corrections and Updates

Nothing yet. This page was published on October 2, 2026.