After GPT-6: OpenAI's next model, the released one and the unannounced one
Checked 2026-10-20: the newest released model in the GPT-6 family is GPT-6.1 Sol, model id gpt-6.1-sol, released by OpenAI on 2026-09-29. OpenAI positions it as near-Astra performance at a lower cost for complex coding, computer use and professional work; tool calling has to go through the Responses API, and reasoning effort supports low, medium (default), high, xhigh and max. The same family also ships GPT-6 Sol (gpt-6-sol) and GPT-6 Luna (gpt-6-luna), both released 2026-09-22, with GPT-6 Astra (gpt-6-astra) as the flagship. On 2026-09-08, in the post announcing its Navier-Stokes Millennium Prize result, OpenAI wrote that it used "an internal model that is significantly more capable than GPT-6 Astra", and that "since August 28 we have been training a new internal model... This model's training is ongoing and its performance continues to improve." So the next model's existence is officially confirmed, not rumoured. That post contains no name, no API model id, no release date and no price; the official post of 2026-09-21 still gives no name or date, and GPT-6 Sol and GPT-6 Luna, released 2026-09-22, are not that internal model. As of 2026-10-20 this page separates the official wording, the checkable timeline, and the preparation that is actually worth doing today.
Updated 2026-10-20
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
The internal model's settled points (each with official wording)
Training start date
Official wording: "Since August 28 we have been training a new internal model". That is the only exact date OpenAI has given for this model, and it anchors the whole timeline.
Training has not finished
Official wording: "This model's training is ongoing and its performance continues to improve". Training still running means no callable version exists today - any claim that it "is already available" is not the official line.
Relative to GPT-6 Astra
Official wording: "we used an internal model that is significantly more capable than GPT-6 Astra". That relative phrasing is all OpenAI gave; the chart in the post carries no citable scores, so no "how much stronger" number has an official source.
Nothing published at all
The post contains no name, no API model id, no release date, no price and no context window. Any name you see elsewhere is not official, and none of those names is repeated here.
The released GPT-6.1 Sol and the internal model still in training
GPT-6.1 Sol is the updated Sol model in the GPT-6 family: OpenAI released GPT-6 Sol (gpt-6-sol) and GPT-6 Luna (gpt-6-luna) on the same day, 2026-09-22, and the official models page points from GPT-6 Sol straight to GPT-6.1 Sol as the newer Sol model. Its input modalities are text and image, its output is text, audio and video are not supported, and it supports US and EU data residency, including with Fast and Ultrafast. As for the stronger model OpenAI confirmed is still training and has still not been named: it is an internal OpenAI model, not a product. In the 2026-09-08 post, OpenAI used it (more precisely, a multi-agent system powered by it) to produce a resolution of the Navier-Stokes existence and smoothness problem, together with a formalization in Lean. What OpenAI has published about it is three points - training began 2026-08-28, training is ongoing, it is significantly more capable than GPT-6 Astra - plus one 2026-09-21 development in the timeline above. It is not a point release of GPT-6 Astra and OpenAI has not placed it in any shipped product line; it is not GPT-6 Sol and GPT-6 Luna, released 2026-09-22 - the launch post still says of GPT-6 Astra that it "continues to be our best model across the board". The most common mistake when reading about this topic is treating capability claims in a research post as specifications of an upcoming product. The post describes an internal evaluation, not a launch.
Checkable numbers in the 2026-09-08 post and on the official model page
The official GPT-6.1 Sol price list (checked 2026-10-20) sets these rates: cached input is priced at 5% of the uncached input token rate, and cache writes are billed at 1.25x the uncached input token rate. Prompts with more than 272K input tokens are priced at 2x input and cache rates and 1.5x output for the full request. Across processing tiers, Fast mode is 2x Standard, Batch and Flex are 50% lower than Standard, and Ultrafast mode is 6x Standard. On 2026-10-08 OpenAI added Ultrafast mode for GPT-6.1 Sol in the Responses API, switched on with model: "gpt-6.1-sol" and service_tier: "ultrafast", available to all API users subject to rate limits, with US and EU data residency. The group that produced the Navier-Stokes resolution was made up of concurrent agents. They reached their resolution on Saturday 2026-09-05; Lean formalization and verification then ran on GPT-6 Astra. OpenAI also states plainly that it does not intend to claim the Millennium Prize for the result. The value of that result is not the scale itself but the yardstick it gives you: it describes a one-off internal research effort, not something anyone can reproduce through an API.
Timeline (each entry names the official post or changelog entry behind it)
OpenAI begins training the new internal model, which it says shows "unprecedented" performance on internal benchmarks including mathematics.
The agents produce the Navier-Stokes resolution. Lean formalization and verification then ran on GPT-6 Astra.
On 2026-09-08 OpenAI published the result and said in the same post that training of the internal model continues and is still improving. Its 2026-09-21 post adds further developments, still with no name and no release date. gpt-6-sol, released by OpenAI on 2026-09-22, is a different model, not the model in training here (see /gpt-6-sol-status-tracker).
Confirmed parts and the parts OpenAI has not published
Officially confirmed (quotable from official pages)
From the developer changelog and the model page, checked 2026-10-20: 1. GPT-6.1 Sol was released on 2026-09-29, model id gpt-6.1-sol. 2. OpenAI positions it as near-Astra performance at a lower cost for complex coding and professional work. 3. Tool calling has to go through the Responses API; Chat Completions is supported without tool calling. 4. It supports US and EU data residency, including with Fast and Ultrafast. 5. Multi-agent is in beta. From the 2026-09-08 post: 1. An internal model exists that is significantly more capable than GPT-6 Astra. 2. It has been in training since 2026-08-28 and training is ongoing. 3. A multi-agent system powered by it produced the Navier-Stokes resolution on 2026-09-05. 4. Lean verification ran on GPT-6 Astra. 5. OpenAI does not intend to claim the Millennium Prize.
The stronger internal model: what OpenAI has still not published
Checked 2026-10-20: the newest GPT-6 family entry in the OpenAI changelog is 2026-10-08, which added Ultrafast mode for GPT-6.1 Sol; the generation after that still has no name and no date. For the internal model — name, API model id, release date, price, context window, output cap, whether it ships publicly at all — the official side has given nothing. One thing not to confuse: what OpenAI published on 2026-09-29 with an id, specifications and pricing is a different model, gpt-6.1-sol, and OpenAI has not presented it as the model in training here; it has been callable on QCode since 2026-09-30. OpenAI has published no name for the model in training, and no other label is an official model: writing a community label as if it were an official model is the most common and hardest-to-undo mistake in this topic. gpt-6-sol, officially released by OpenAI on 2026-09-22, is a different model; its specs and pricing are at /gpt-6-sol-status-tracker. If you need one sentence for your team: officially training is confirmed, no date was given.
GPT-6.1 Sol, GPT-6 Astra and the internal model you cannot call
GPT-6.1 Sol / GPT-6 Astra - callable today
GPT-6.1 Sol: model id gpt-6.1-sol, released 2026-09-29, 1,050,000 context and a 128,000 output cap that includes reasoning tokens, list price $2.00 input, $0.10 cached input, $2.50 cache write and $10.00 output per 1M tokens, with reasoning effort starting at low and medium as the default. GPT-6 Astra: released 2026-09-03, API model id gpt-6-astra, 1,050,000 context and a 128,000 output cap that includes reasoning tokens, list price $10 per 1M input and $50 per 1M output. The Lean formalization and verification above ran on this model, so it is already capable of that class of work.
The internal model - out of reach
The only relative capability phrase OpenAI gave is "significantly more capable"; its 2026-09-21 development is in the timeline above. No id means no way to call it; no date means no way to schedule around it. Its practical implication for you today is to write code where the model name is one line of config, rather than to wait.
The preparation that is actually worth doing
Checked 2026-10-20: the official migration notes list the parameters to change when you move to GPT-6.1 Sol - when reasoning effort is not "none", remove temperature, top_p and top_logprobs; for Chat Completions, also remove logprobs; and from include, remove message.output_text.logprobs. Requests that used minimal should start at low instead. Lift the model name out of your code and into configuration or an environment variable, so switching costs one line. Every OpenAI model shipped this year (the three GPT-5.6 tiers, GPT-6 Astra) speaks the same Responses API; migration friction comes from parameters rather than from the interface. Astra drops temperature, top_p and top_logprobs, has no "none" reasoning tier so you start at "low", and requires tool calls to go through Responses. Concentrate those differences in one adapter layer and the next model - whatever it is called, whenever it lands - costs you a config change rather than a code change. This is work you can finish today, and it does not depend on anything unreleased.
How to use it on QCode
One QCode key calls gpt-6.1-sol, gpt-6-sol, gpt-5.6-sol, gpt-5.6-terra and gpt-6-astra (status as of 2026-10-20 on /models); switching is a model id change, with no new endpoint and no new auth. QCode bills per token, and per-model rates are listed on the /models page. OpenAI released gpt-6.1-sol on 2026-09-29; it has been callable on QCode since 2026-09-30. QCode does not currently support gpt-6-luna or gpt-5.6-luna (both are temporarily offline on QCode); please use another GPT model, such as gpt-6.1-sol, gpt-6-sol or gpt-5.6-terra. That means you can build the adapter layer described above on gpt-6.1-sol or gpt-6-sol today and measure your cost baseline now. When the next model is actually released publicly, the work left for you is adding one more model id. We will not promise a launch date for any unreleased model - OpenAI has not given one, so we certainly should not. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days (top-ups start at $5).
Frequently asked questions
What is OpenAI's next model called?
The released answer is GPT-6.1 Sol (gpt-6.1-sol), released 2026-09-29. For the stronger internal model OpenAI confirmed is still training, no name has been published: the 2026-09-08 post refers to it only as an "internal model" throughout. Any specific name is unofficial, including the codenames circulating in the community and including "GPT-7", which is simply extrapolated from the numbering.
When will it be released?
The released GPT-6.1 Sol came out on 2026-09-29. For the internal model still in training, OpenAI's wording is that training is ongoing and still improving, so no release date has been published. That wording itself rules out any firm "soon" claim - a model that is still training has no release schedule. Any text with a specific date comes from somewhere other than the official source.
How much stronger is it than GPT-6 Astra?
OpenAI wrote only "significantly more capable". The chart in the post plots pass rate against test-time compute and carries no citable numbers, so no "how many percent stronger" figure has an official source. The one checkable capability data point is that a multi-agent system powered by it resolved Navier-Stokes. The released GPT-6.1 Sol does have an official comparison: near-Astra performance for complex work at a lower cost.
Can I use it now?
The internal model is not callable: OpenAI explicitly calls it an internal model and says training is ongoing, and with no API model id there is no way to call it. The newest model you can call today is GPT-6.1 Sol (gpt-6.1-sol, released 2026-09-29); the previous flagship GPT-6 Astra was released 2026-09-03.
What does the Navier-Stokes result actually prove about its strength?
It shows the system can produce a Lean-verifiable proof for a problem open for roughly 90 years, which is hard evidence. Three caveats belong with it: that result came from a large group of concurrent agents working together, not from one conversation; the verification step ran on GPT-6 Astra; and OpenAI itself declines to claim the Millennium Prize and calls the result "a snapshot in time" rather than a culmination.
Should I wait for it, or move to GPT-6 Astra now?
Waiting for something with no date is not an executable plan. The better move is to lower your switching cost now: put the model id in config and concentrate parameter differences in one adapter layer. Then whenever the next model lands, switching costs one line. The usage patterns and cost baseline you build on gpt-6.1-sol or GPT-6 Astra today stay valid either way.
Sources
Official quotes and numbers on this page come from OpenAI's official blog posts of 2026-09-08 (openai.com/index/navier-stokes-solution/), 2026-09-21 (openai.com/index/advisory-group-on-mathematics-and-ai/) and 2026-09-06 by Chief Scientist Jakub Pachocki (openai.com/index/an-alien-mind/), plus the launch post of 2026-09-22 (openai.com/index/introducing-gpt-6-sol-and-luna/) and the OpenAI developer changelog (developers.openai.com/api/docs/changelog). Specifications, pricing and interface notes for GPT-6.1 Sol, GPT-6 Sol, GPT-6 Luna and GPT-6 Astra come from the official model pages and models overview (developers.openai.com/api/docs/models) and the Using GPT-6 guide. QCode's model list and billing wording come from QCode's models page and pay-as-you-go page. No secondary retelling is cited and no codename without an official source is repeated. The update date appears in the page footer.
Lower your migration cost first, then wait for the new model
One key calls GPT-6.1 Sol, GPT-6 Astra and the three GPT-5.6 tiers; switching is a model id change. When the next model lands, your job is adding one line.
Related pages
Complete GPT-6 guide
GPT-6 Astra specification, pricing and practical usage.
GPT-6 Astra migration guide
The four parameter changes needed when moving from GPT-5.6 to Astra.
GPT-6 Astra quota burn
How Astra consumes quota in Codex, with measured community observations.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
For the internal model that has not been released, every "officially confirmed" item above can be checked verbatim in the official posts listed on this page; name, release date, price and API model id have not been published, and no speculation is made here and no launch date is promised for any unreleased model. That model is still unreleased; what OpenAI released on 2026-09-29 is the different model gpt-6.1-sol, callable on QCode since 2026-09-30.