Information correct as of 2 October 2026. OpenAI can change prices, plan allowances and where dots are available, so check your own account before you commit budget.
OpenAI used DevDay 2026 to make two arguments at once. The first is that an AI agent is now something a business subscribes to rather than something it builds: dots, always-on agents inside ChatGPT, arrived with a cloud computer each and a place on the Pro and Business Premium price lists. The second is that near-flagship model quality now costs $2 per million input tokens and $10 per million output, the price of GPT-6.1 Sol and of Claude Sonnet 5.5 alike. For a UK business the catch sits in the small print. Dots on a personal Pro plan are not available in the UK, and the fastest tier and the agents platform both keep their data in the United States.
DevDay 2026 ran on Tuesday 29 September at Fort Mason in San Francisco. OpenAI’s recap lists 25 items, “more than 20 major announcements” in its own words, and three carry the day. Dots run on GPT-6 Astra and the first one is included with a Pro or Business Premium plan. GPT-6.1 Sol is an upgrade to GPT-6 Sol that OpenAI says “nearly matches GPT-6 Astra’s intelligence” at a fifth of Astra’s input and output prices. Pro 500 is a new $500-a-month ChatGPT tier with Ultrafast built in. OpenAI’s eligibility page limits dots on Pro plans to “users over 18 outside the European Economic Area, United Kingdom, and Switzerland”, while Business Premium and Enterprise workspaces are “rolling out worldwide”.
The risk for a UK team is not that it misses an agent it cannot buy yet. It is that the economics of its model spend changed on 29 September, and that a pilot started on the wrong plan, or in the wrong region, will stall for reasons that were printed on OpenAI’s help pages from the first day.
Dots are the product, and UK Pro plans are not in the first wave
OpenAI’s launch post calls dots “remarkably capable, always-on agents built to handle everything”, “powered by GPT-6 Astra”, each with “their own cloud computer”. The documentation is plainer: a dot is “an always-on agent that keeps work moving and asks for your input when a decision needs you”. You create one in the ChatGPT desktop app or a desktop browser, then message it from ChatGPT, Slack or Teams; the mobile app follows once its supporting update reaches your account. Through plugins a dot can reach over 4,000 apps. By every visible sign it is the “o” assistant that TestingCatalog reported the week before: the Pro pricing card that read “o, your always-on assistant” now reads “Dot, your always-on agent”, though OpenAI has not said so.
The commercial terms at launch are simple. The first dot is included in a Pro or Business Premium plan “at no extra cost”, conversations with it do not count towards ChatGPT usage limits, and OpenAI’s help centre says that “in the future, you’ll be able to add more dots, and scale the output of each dot by either increasing its speed or the total amount of work it can take on per month”. No price has been published for either.
Read the controls before an agent is let loose on company accounts. A dot works on its own cloud computer, and your own machine stays separate unless you connect it. OpenAI says saved passwords are used “without exposing them to the model”, proactive research tools are “restricted to be read-only”, and custom rules run from “Take action without asking” to “Hand off to you”. Business, Enterprise and Edu content is not used for training by default.
Can UK businesses use dots yet?
On a Pro plan, no. The eligibility page, checked 2 October 2026, reads: “Pro 100, Pro 200, and Pro 500: For users over 18 outside the European Economic Area, United Kingdom, and Switzerland.” The help centre puts it the other way round: “Dots are rolling out today in ChatGPT to Pro users in markets excluding the European Economic Area, Switzerland, and the UK.” OpenAI gives no reason and no date, and it is not alone in holding Europe back: Meta’s Muse launched in the United States only, as our Meta Connect recap sets out.
On a workspace plan, yes, in principle. The same page says “Business Premium: Rolling out worldwide” and “Enterprise: Rolling out worldwide. Dots are off by default and must be enabled by a workspace administrator.” Business Premium is the seat OpenAI introduced on 10 August at $125 per user per month, or $100 a month on an annual plan; our guide to ChatGPT Work covers the workspace plans. So a UK company that wants a dot pilot buys Premium seats for the people who will run it, or asks its Enterprise administrator to switch dots on, and then waits, because OpenAI says “Dots are rolling out gradually. You may not see dots immediately, even if your plan is eligible.” One more caveat for anyone measuring a pilot: the documentation promises “extended limits for the first month after launch”, so what a dot gets through in October is no guide to November.
Near-flagship work now costs $2 in and $10 out
GPT-6.1 Sol is the model launch of the day, and its price is the news. Input and output are unchanged from GPT-6 Sol at $2 and $10 per million tokens, and cached input is halved to $0.10, “50% less than GPT-6 Sol’s cached input pricing” in OpenAI’s words. That puts it level with Claude Sonnet 5.5 on headline price and at a fifth of GPT-6 Astra, as the table shows. OpenAI’s own benchmarks are vendor claims until someone independent re-runs them, but they are specific: on Zapier’s AutomationBench the model scores “2.2 percentage points above Opus 5.5 at medium reasoning effort, at roughly a third of the cost”, and on DeepSWE v1.1 it “matches GPT-6 Astra at roughly one-fifth of the cost”.
If the table extends beyond the screen, scroll sideways to view all columns.
| Model | Input | Output | Cache read |
|---|---|---|---|
GPT-6.1 Sol (gpt-6.1-sol) |
$2.00 | $10.00 | $0.10 |
GPT-6 Astra (gpt-6-astra) |
$10.00 | $50.00 | $1.00 |
| Claude Sonnet 5.5 | $2.00 | $10.00 | $0.20 |
| Claude Opus 5.5 | $4.00 | $20.00 | $0.20 |
| Gemini 3.8 Flash | $0.75 until 31 December 2026, then $1.50 | $3.75 until 31 December 2026, then $7.50 | $0.075 until 31 December 2026, then $0.15 |
There is an independent reading too. Artificial Analysis scores GPT-6.1 Sol at 52 on its Intelligence Index v4.3.2, one point behind GPT-6 Astra on 53 and four behind Claude Sonnet 5.5 on 56, at $0.72 per Index task against $3.26 for Astra and $7.62 for Sonnet 5.5, which produced 410 million output tokens across the Index against Sol’s 67 million. Those are different effort settings, read on 1 October 2026 from an index that is re-run, so treat the gap as the point rather than the decimals.
GPT-6.1 Sol is in the API, and in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu, but “not yet available in Chat”. Procurement teams should note one line in OpenAI’s system card addendum, which treats the model “as Critical in cybersecurity and High for Biological and Chemical capability” and gives it “the same safeguards stack as GPT-6 Astra”. The cheaper model carries the flagship’s safeguards.
Ultrafast and Pro 500 come with residency small print
Ultrafast is OpenAI’s fastest service tier, “up to 8× faster token generation (300 tokens per second) in Codex and up to 6x in the API”, and since 29 September it has been “broadly available for GPT-6 Astra” in the API at $60 per million input tokens and $300 output, six times the Standard rate. GPT-6.1 Sol Ultrafast is “coming soon” with no published price. The line UK firms should read twice is in the Ultrafast guide: the tier “supports US data residency and global processing only. It does not support EU or other non-US regional processing endpoints”. The Agents API, which gained computer use at DevDay and remains a beta, says the same of itself: it “supports data residency only in the United States and does not support Zero Data Retention”. A company that has promised EU customers regional processing cannot use either for that work.
Pro 500 is the plan built around Ultrafast. OpenAI’s help page lists three monthly Pro tiers in US dollars, Pro 100 at $100, Pro 200 at $200 and Pro 500 at $500, and says “Ultrafast is available only on Pro 500”. The quieter change is to Pro 200: it “is also available for new subscriptions again”, but new subscriptions “include a lower usage allowance”, and existing subscribers keep their previous allowance only “through Oct 29, 2026”. The dollar figures are US list prices. ChatGPT bills UK subscribers in pounds at a localised price shown at checkout, per OpenAI’s multi-currency billing article, so check the pound figure before it goes in a budget; API credits are sold in dollars only.
What OpenAI did not ship
Two models expected before the event stayed off the stage. AFP, reported by Malay Mail, quoted Saachi Jain, OpenAI’s head of safety systems, saying that Astra 6.1 “didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done”. OpenAI’s own pages do not mention a GPT-6.1 Astra, so treat the name and the timing as unconfirmed. GPT-6 Cyber, which Fortune reported could be previewed at DevDay, is not on the recap; Fortune put it “in the coming weeks”, and OpenAI has said nothing either way.
The context is a fortnight of hard safety news that the recap does not mention. On 28 September OpenAI published “How we will do better for Australia”, saying it “should have shared preliminary findings sooner” after its models, “during internal training and evaluation”, had “accessed Australian government websites in ways they were not authorised to”. The same week the UK AI Security Institute reported that GPT-6 Astra carried out unsanctioned supply-chain attacks in simulated evaluations at a higher rate than GPT-5.6 Sol. What did ship, GPT-6.1 Sol, carries Astra’s safeguards.
A platform assembled in seven weeks
DevDay did not start from nothing. Between 13 August and 22 September OpenAI previewed Ultrafast, released GPT-6 Astra, launched the Agents API in public beta, put GPT-Live-1 in the API, released GPT-6 Sol and Luna at half the GPT-5.6 promotional rates and rebuilt prompt caching with a 90% discount on cached reads. DevDay productised the pieces. The Agents API gained computer use and an AWS home in Bedrock Managed Agents, now in preview, and GPT-6.1 Sol took GPT-6 Sol’s place on the pricing page’s flagship table a week after Sol launched.
The pattern matches last year. DevDay 2025 turned the Apps SDK, AgentKit and Codex into products; a year on, Agent Builder, the centre of that AgentKit launch, sits under “Legacy APIs” with a migration guide. Expect the same half-life.
What should UK marketing teams do next?
DevDay changes the numbers more than it changes the plan. Five checks, in this order:
- Move GPT-6 Sol work to GPT-6.1 Sol. Same input and output prices, cheaper cached input, better scores on every capability benchmark OpenAI published, and GPT-6 Sol has already left the flagship pricing table. Our GPT-6.1 Sol guide has the full price, benchmark and migration detail.
- Test Astra and Opus workloads on the $2 models. If a workflow runs on GPT-6 Astra or Claude Opus 5.5 and is not frontier-hard, run it on GPT-6.1 Sol and Claude Sonnet 5.5 before the next invoice, and compare cost per task rather than price per token; our Claude Sonnet 5.5 guide explains why the two diverge.
- Run any dot pilot on Business Premium, not Pro. Budget $125 a seat at US list price, check the pound figure at checkout, and discount the first month’s results for the launch allowance.
- Write the residency exclusions into the data protection assessment. Ultrafast and the Agents API are US-only on residency, regional endpoints carry a 10% uplift, and a pilot that promises EU customers regional processing cannot use either.
- Prepare the content a dot will read. A dot researching suppliers for its owner browses on their behalf; structured, specific, verifiable pages serve it better than a homepage, and our guide to ChatGPT, AI search and AEO sets out how.
What to watch next
The first thing to watch is a UK date for dots on Pro; OpenAI has given none. After that: a price for GPT-6.1 Sol Ultrafast and for a second dot, documentation for the Decisions API, which is “available in limited preview today with a broad release planned in the coming days”, and whether GPT-6 Cyber appears. As of 2 October 2026 OpenAI’s newsroom and API changelog carried nothing on any of them. Closer to home, the London DevDay Exchange is on Tuesday 3 November at 2pm, one of eight follow-on events, and OpenAI’s events page says applications “are currently being reviewed, but please still feel free to apply”.
The real lesson from OpenAI DevDay 2026
DevDay 2026 was a pricing event dressed as a product launch. The agent was the headline, but the number that changes a UK budget is $2 in and $10 out for work that, a month earlier, cost five times as much on OpenAI’s own flagship.
For UK teams the lesson is to separate what can be bought today from what cannot. The model is available now, in the API and in ChatGPT Work, and the test is cheap. The agent arrives on workspace plans first and on UK personal plans at a date OpenAI has not given, which leaves time to put a pilot on the right plan, in the right region, before anyone is asked to measure it.
Last year’s recap, OpenAI DevDay 2025 for digital marketing, shows how quickly this platform moves on. If you want a second opinion on your model spend, a dot pilot on Business Premium, or how your content reads to an agent, talk to AIWIZ.
Frequently asked questions
When and where was OpenAI DevDay 2026 held?
Tuesday 29 September 2026 at Fort Mason in San Francisco, with Sam Altman’s keynote at 10am Pacific, 6pm UK time. The full 53-minute keynote is on OpenAI’s YouTube channel, and OpenAI’s written recap at openai.com/index/devday-2026-recap lists all 25 announcements.
Are dots available in the UK?
Not on a Pro plan. OpenAI’s eligibility page, checked 2 October 2026, limits dots on Pro 100, Pro 200 and Pro 500 to users over 18 outside the European Economic Area, United Kingdom and Switzerland. Business Premium and Enterprise workspaces are rolling out worldwide, with Enterprise dots off until an administrator enables them.
What does GPT-6.1 Sol cost?
$2 per million input tokens, $0.10 for cached input and $10 per million output tokens on the Standard tier for prompts up to 272,000 input tokens. Input and output match GPT-6 Sol and Claude Sonnet 5.5; cached input is half GPT-6 Sol’s $0.20. Long-context and Fast mode rates are higher. Checked on OpenAI’s pricing page on 2 October 2026.
What did OpenAI not announce at DevDay 2026?
A GPT-6.1 Astra, which OpenAI’s head of safety systems told AFP had not met its safety bar, and GPT-6 Cyber, which Fortune had reported could be previewed. Neither is on OpenAI’s recap, and OpenAI has confirmed neither. GPT-6.1 Sol Ultrafast and the Decisions API were announced without published prices or documentation.
When is the London DevDay Exchange?
Tuesday 3 November 2026 at 2pm, one of eight follow-on events after Bengaluru, Tokyo, Seoul, Berlin and Paris and before São Paulo and Mexico City. OpenAI’s events page said applications were being reviewed but still open, checked 2 October 2026.