OpenAI DevDay 2026: what OpenAI announced, and what it still hasn't published
OpenAI held DevDay 2026 at Fort Mason in San Francisco on 2026-09-29, calling it its biggest yet with more than 20 major announcements. The release that touches developers most directly is the new API model gpt-6.1-sol: standard rates of $2 input, $0.10 cached input, $2.50 cache writes and $10 output per million tokens. As of 2026-09-30 QCode does not list gpt-6.1-sol as a callable model.
Updated 2026-09-30
Four things DevDay 2026 is easy to read wrong
The released model is gpt-6.1-sol, not the cancelled GPT-6.1 Astra
Outlets reported on 2026-09-28 that GPT-6.1 Astra failed OpenAI's internal safety bar and would not ship; the model OpenAI released the next day is a different id, gpt-6.1-sol. OpenAI never says the two are the same thing, and neither does this page.
GPT-6 Sol's pricing and specification did not change at DevDay
The gpt-6-sol row still reads $2.00 / $0.20 / $2.50 / $10.00, and the model page still lists 1,050,000 / 922,000 / 128,000. What is new is a same-priced gpt-6.1-sol with a different cached-input convention: 5% instead of 10%.
Not in Chat: 6.1 Sol sits in Work and Codex, like Sol and Luna
The official Codex documentation states that in ChatGPT, GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna are available in Work and Codex and are not available in Chat. Launch covers Plus, Pro, Business, Enterprise and Edu; Free and Go are not included at launch, and Enterprise and Edu keep it off until an administrator enables it.
Benchmarks stay out: OpenAI says Astra scores highest; we do not transcribe numbers
The launch post carries OpenAI's own score table and states that Astra still achieves the highest score among the models tested and should be used for the hardest research tasks. This page records that claim as a claim and transcribes no scores; for prices and specifications see the comparison above.
What DevDay 2026 was actually for
This is a developer conference: the official recap states «developers can directly launch new native experiences to our collective 1.2B weekly users» (the 1.2B figure is OpenAI's own claim). Of the list it published, three items bear directly on the API: the new model gpt-6.1-sol, computer use in the Agents API, and the Ultrafast tier in the Responses API. The rest are ChatGPT product surfaces (ChatGPT Space, Pages, collaborative slides, @ChatGPT, the Meetings plugin, Sign in with ChatGPT, Marketplace) and a subscription tier (Pro 500).
2026-09-29: OpenAI releases gpt-6.1-sol
OpenAI's API changelog lists, under 2026-09-29, «Released GPT-6.1 Sol (gpt-6.1-sol)», positioned for complex coding and professional work at a lower cost than GPT-6 Astra. The launch post calls it an upgrade on GPT-6 Sol, «near-Astra intelligence … at one-fifth of Astra's standard API input and output token prices» — and the official table does give $2 against $10 for input, $10 against $50 for output. OpenAI also states Astra still scores highest among the models tested; this page does not transcribe any benchmark numbers.
Release day and what follows
DevDay 2026 itself: the 2026-09-29 block of the official API changelog lists the gpt-6.1-sol release and three API changes, and the launch post and recap carry the same day. The official schedule puts the opening keynote at 10:00 a.m. Pacific Time.
The date this page was checked, 2026-09-30. On the QCode side, the 30-day usage table shows no calls for gpt-6.1-sol, while gpt-6-sol, gpt-6-luna and gpt-6-astra are still being called. What is available on QCode today is not limited to those: the site's /models page is the authority, and this page does not enumerate it.
As of 2026-09-30 this is OpenAI's wording, not a date: Ultrafast for GPT-6.1 Sol and the broad release of the Decisions API are both «in the coming days», collaborative slides are «in the coming weeks», the Private Inference preview is «this fall». We keep the wording as published and do not turn it into calendar dates.
Layers of sourcing: official text / vendor claim / our arithmetic / site measurement
Published by OpenAI (verbatim source)
Officially announced. Model gpt-6.1-sol: 1,050,000 context, 922,000 maximum input, 128,000 maximum output, knowledge cutoff 2026-04-30, input modalities text and image, reasoning efforts low / medium (default) / high / xhigh / max, with none and minimal unsupported. Standard rates per million tokens: $2 input, $0.10 cached input, $2.50 cache writes, $10 output; requests above 272K input tokens are billed at 2x input and cache rates and 1.5x output for the whole request; cache writes are 1.25x; Fast is 2x Standard while Batch and Flex are 50% lower; regional processing adds 10% where available. Three API changes landed the same day: computer use in the Agents API with an OpenAI-hosted browser, Ultrafast for GPT-6 Astra in the Responses API, and Multi-agent in beta for the gpt-6.1-sol model.
Not published (left blank, no guessing)
Not published as of 2026-09-30, and this page carries no rumour and no leak: a retirement date for gpt-6.1-sol (the official deprecations page has no entry for it), the Ultrafast rate (the official Ultrafast table lists only gpt-6-astra), which markets count as dots' «eligible markets», the dollar prices of dots and of the Pro 500 plan, and dates or ticket prices for the eight DevDay Exchanges cities. The only timing OpenAI gives is wording: broad release of Ultrafast and the Decisions API «in the coming days», collaborative slides «in the coming weeks», the Private Inference preview «this fall». We do not convert those into dates.
Two rates published the same day: the new 6.1 Sol and the current Astra
gpt-6.1-sol official standard rates
The Standard table on the official pricing page: $2 input, $0.10 cached input, $2.50 cache writes, $10 output; the long-context tier (input above 272K) is $4.00, $0.20, $5.00, $15.00. Cached input, per OpenAI, is priced at 5% of the uncached input rate.
For comparison: gpt-6-astra official standard rates
The Astra row of the same table: $10.00 input, $1.00 cached input, $12.50 cache writes, $50.00 output. By that table the input and output of 6.1 Sol come to exactly one fifth of Astra's, while the cached-input convention drops from 10% to 5%.
Three things a developer can act on today
First, for coding work that wants near-Astra behaviour at lower cost, the official guide describes gpt-6.1-sol as balancing intelligence and cost, and requires tool calling through the Responses API. Second, for browser-shaped agents, OpenAI now ships computer use in the Agents API: the browser runs in an OpenAI-hosted environment while your application handles site access approvals and sign-in. Third, when choosing inside the GPT-6 family, the official migration checklist starts by setting model to gpt-6-astra, gpt-6.1-sol or gpt-6-luna and then checking each item; OpenAI states Astra remains the right choice for the hardest research tasks.
The QCode side: DevDay's new model is not with us today
As of 2026-09-30 this page does not list gpt-6.1-sol as callable on QCode and gives it no price. Models from OpenAI available on QCode today include gpt-6-astra, gpt-6-sol and gpt-6-luna; the Anthropic side includes claude-opus-5-5 and claude-sonnet-5 — the authority for the full list is the site's /models page. dots, Pro 500, Codex cloud and Marketplace are subscriptions and agents hosted by OpenAI; QCode does not sell those, and they appear here as background only.
Questions
What date was DevDay 2026?
The official schedule puts it on Tuesday 2026-09-29 at Fort Mason in San Francisco, with Sam Altman's opening keynote at 10:00 a.m. and a closing session at 4:00 p.m. Registration is $650 for invitees; the keynote is livestreamed free.
What is the new model called and what does it cost?
The official API model id is gpt-6.1-sol. Standard rates per million tokens are $2 input, $0.10 cached input, $2.50 cache writes and $10 output. Requests above 272K input tokens bill at 2x input and cache and 1.5x output for the whole request; Fast is 2x Standard; Batch and Flex are 50% lower; regional processing adds 10% where available.
Is it the same model as GPT-6.1 Astra?
This page does not say that. Reporting on 2026-09-28 concerns GPT-6.1 Astra, which OpenAI decided not to release; the model released on 2026-09-29 is gpt-6.1-sol. OpenAI never links the two, so the honest boundary here is «not stated».
Can I call gpt-6.1-sol on QCode?
As of 2026-09-30 this page does not list it as callable and gives it no price. Models from OpenAI available on QCode today include gpt-6-astra, gpt-6-sol and gpt-6-luna; whether and when gpt-6.1-sol joins is a question for the live /models page.
What are dots, and what do they have to do with the API?
dots are a class of always-on agents released the same day. OpenAI describes them as powered by GPT-6 Astra, connecting to more than 4,000 apps with their own cloud computer and browser, rolling out to Pro, Business Premium and Enterprise in «eligible markets» (no market list is published). It is a ChatGPT product surface, not a model id you can put in code.
OpenAI says 20-plus announcements — why does this page list fewer?
This page keeps only what directly affects people writing code and what OpenAI published a sentence for. The rest — ChatGPT Space, Pages, collaborative slides, @ChatGPT, the Meetings plugin, Sign in with ChatGPT, Marketplace, Bedrock Managed Agents, plugin extensions — appears here as a list, because they are product surfaces rather than interfaces and models.
Sources
Seven official OpenAI pages: the API changelog, the gpt-6.1-sol model page, pricing, deprecations, the Codex model documentation, the latest-model guide and devday.openai.com. Three launch posts (introducing-gpt-6-1-sol, devday-2026-recap, introducing-dots) were unreachable directly, so we took their 2026-09-29 Wayback snapshots plus openai.com/news/rss.xml. Archives were captured 2026-09-30 02:31–02:44 Beijing time. The site-side claim uses only the 30-day usage record re-checked on 2026-09-30.
Want the GPT-6 family's rates and specs? Start with these three
gpt-6-astra, gpt-6-sol, gpt-6-luna and claude-opus-5-5 share one QCode key; the gpt-6.1-sol released at DevDay is not here as of 2026-09-30.
Related pages
GPT-6 Sol / Luna pricing tracker
The official table, four billing rules and four worked examples. DevDay changed nothing in GPT-6 Sol's numbers, and the comparison above is drawn from that same table.
GPT-6.1 Astra release cancelled
What the outlets reported, what OpenAI actually said, and what is still callable today — a different model id from the gpt-6.1-sol discussed here.
GPT-6 Sol status tracker
Sol's own release and status timeline. After DevDay it remains the coding tier QCode serves from OpenAI.
This page carries only what OpenAI's own pages state verbatim: capability claims and multiples are labelled as OpenAI's own statements, wording such as «in the coming days / weeks» and «this fall» is kept as published rather than converted into dates, and no benchmark numbers are transcribed. QCode is independent of OpenAI, and this page makes no availability claim for gpt-6.1-sol; the site's usage figures are an observation of 2026-09-30 only and never a service level.