Platin Digitalplatindigital.
English
Let's talk

AI news15 min read

OpenAI DevDay 2026: Every Announcement, and What It Means for Your Business

OpenAI DevDay 2026 brought dots, GPT-6.1 Sol, Ultrafast and the Agents API. What was announced, who can use it now, and what really matters for a small business.

Shahaf PlatinPlatin Digital · OpenAI Select Partner

OpenAI DevDay 2026: Every Announcement, and What It Means for Your Business
AI-generated illustration

Tuesday, September 29, 8:00 PM Israel time

It's ten in the morning in San Francisco, and the DevDay 2026 keynote is about to begin. The day before, OpenAI announced that it would not release GPT-6.1 Astra, because in safety testing it tended to step outside the limits of its task and its permissions. Everyone was waiting to see what would take its place.

What took its place isn't a smarter model. It's an employee who never goes home.

OpenAI DevDay 2026 was the conference where the central question changed. Less "which model is smartest," more "who keeps working when you close your laptop, and who decides what it's allowed to do." OpenAI described more than 20 announcements. We've gathered the ones that matter, checked every detail against official sources, and added the question you actually care about: what does this mean for a business?

In short

  • dots: standing agents inside ChatGPT, with their own computer and browser in the cloud. This is the headline announcement of DevDay 2026, and it is rolling out gradually to Pro, Business Premium and Enterprise subscribers in select markets.
  • GPT-6.1 Sol: a new API model that, according to OpenAI, comes close to GPT-6 Astra's performance at one fifth of the price per token.
  • Ultrafast and the Agents API: a pricier speed tier, and a beta infrastructure where developers run agents that OpenAI manages for them.
  • For business owners, the important change isn't another model but agents that keep working in the background, with clear rules about when they ask you before acting.

An engine, a driver, and who holds the wheel

To avoid getting lost among 20 announcements, think of a car.

The models are the engine. GPT-6.1 Sol and GPT-6 Astra determine how powerful and how fast it goes, and how much fuel it burns.

The agents are the driver. dots and the Agents API are the part that takes the engine and drives it somewhere specific: opens a website, reads an email, prepares a document.

The rules and approvals are the hands on the wheel. And this is exactly what makes this year's DevDay interesting. OpenAI gave up a stronger engine, GPT-6.1 Astra, because its driver didn't stop when it should have. That same week it launched drivers that work around the clock, with explicit rules about what they may do on their own.

What was announced at DevDay 2026: the full table

AnnouncementWhat it isAvailabilitySource
dotsA standing agent inside ChatGPT, with a cloud computer and browser and connections to appsRolling out gradually to Pro, Business Premium and Enterprise in select markets. The first dot is included in the subscriptionOpenAI announcement, details per MacRumors and BGR
GPT-6.1 SolA model for coding, computer use and professional work, cheaper than AstraIn the API from September 29Official changelog
UltrafastThe highest speed tier in the API, at a higher priceFor GPT-6 Astra in the API, with low rate limits. For Sol laterOfficial documentation
Agents APIManaged infrastructure for running agents: sessions, workspace, tools and sub-agentsBetaOfficial documentation
Computer use in the Agents APIAn agent that works in a browser hosted by OpenAIFrom September 29Official changelog
Codex CloudCoding tasks that keep running in the cloud even when your computer is offDocumented on the Codex help siteOfficial documentation
ChatGPT Space and PagesA shared workspace for a team and its dots, and documents edited togetherAvailability details weren't published in the sources we checkedPer BGR and Engadget
Sign in with ChatGPTSign in to other apps with your ChatGPT account and use your subscription quota in them16 partners at launch, per BGROfficial documentation
MCP EventsA connected app can "wake up" ChatGPT when something happens on its sideFor plugin developersOfficial documentation
ZDR with Private Safety ProcessingAutomated safety checks without OpenAI storing the contentFor organizations approved for Zero Data RetentionOfficial documentation
Pro 500A new, pricier Pro tier, and the return of Pro 200 for new subscribersPer Engadget and BGRPress reports

dots: an employee who never goes home

This is the announcement everyone will be talking about. A dot is an agent that lives inside ChatGPT and doesn't disappear when the conversation ends. According to OpenAI, it is a proactive helper that keeps working on projects and tasks between conversations.

According to MacRumors and BGR, each dot runs on GPT-6 Astra, gets its own computer and browser in the cloud, and can connect to more than 4,000 apps through ChatGPT plugins. You talk to it in ChatGPT, Slack or Microsoft Teams, and voice conversations are possible too. It remembers context across channels, so you can start a task in ChatGPT and continue it in a Slack thread.

The interesting part is the limits. According to BGR, when you're not working with it, the dot does "proactive research": it goes through the connected apps using read-only tools and looks for what might be worth doing next. At this stage it doesn't send messages, doesn't edit and doesn't take control of the computer. On top of that there are Custom Rules, where you define for each action whether it's allowed, requires approval or is blocked. Sensitive actions, like changing a password, stay with you.

How a dot works, as OpenAI presented it

  1. You set a goaland connect the apps it needs
  2. Background researchRead-only access in the connected apps
  3. Prepares an actionand checks it against the rules you set
  4. Acts or asksPer Custom Rules: allowed, with your approval, or blocked

In gold: steps where you are involved. Based on BGR and MacRumors reports from the conference.

Who can get it? According to MacRumors, dots are rolling out gradually to Pro, Business Premium and Enterprise plans in select markets, and other users will get access later. The first dot is included in the subscription. Conversations with it don't count toward ChatGPT's usage limits, but tasks it runs do. According to BGR, in Enterprise organizations the system administrator has to enable the option. Text-message conversations and teams of multiple dots will come later. If you're in Israel, check your account to see whether the option already appears.

What does this mean for a small business owner? For the first time, an agent that works in the background is available as an off-the-shelf product, inside ChatGPT, with nothing to set up. For some businesses that will be enough. But watch out for two things: it works through apps that have a ChatGPT plugin, and the subscription it requires is on the pricier tiers. If your work lives in a WordPress site, a Google Ads account and your own systems, you still need to check what actually connects and with what permission. We covered this in our guide to an AI agent for your business.

GPT-6.1 Sol: almost Astra, at one fifth of the price

GPT-6.1 Sol arrived in the API on September 29. According to OpenAI it comes close to GPT-6 Astra's performance in coding, computer use and professional work, and its price per token is one fifth of Astra's. What matters for business owners: a cheaper model at a close level means tasks that were too expensive for an agent, like a long pass over a website or large documents, may become worth it.

$2 / $10GPT-6.1 Sol price per million tokens, input and output, on the standard tier
$10 / $50The equivalent GPT-6 Astra price. Hence "one fifth of the price"
1,050,000tokens in GPT-6.1 Sol's context window

Source: OpenAI's developer model page and pricing page, September 30, 2026. The price applies to requests of up to 272 thousand input tokens. Beyond that the price is higher.

But it's worth being honest: "close to Astra" is OpenAI's wording, and the real test is on your tasks. OpenAI itself recommends in its documentation that you compare Sol against Astra on your own tasks before choosing.

Ultrafast: speed, at full price

Ultrafast is the fastest service tier in the API. In its documentation OpenAI writes that it reaches up to 8 times the speed of the standard tier. For now it's available for GPT-6 Astra to all API users, with low rate limits, and support for Sol will come later. It doesn't support processing in the European Union. According to BGR, in ChatGPT it is currently available only on the Pro 500 and Enterprise tiers.

The price: according to the official pricing table, Astra on Ultrafast costs 6 times the standard tier, $60 per million input tokens and $300 for output. For most small businesses this isn't relevant. An agent that prepares a weekly report doesn't need to answer in half a second. This is meant for services where every second of waiting costs you a customer.

Agents API and Computer use: the infrastructure under the agents

This is the technical part of DevDay, and it's less flashy but important. The Agents API, which is in beta, gives developers access to the same infrastructure that runs Codex, through an API that OpenAI manages. OpenAI takes care of sessions, coordination and recovery from failures. The developer supplies the tools and chooses the workspace. The agent can run code, edit files, connect to MCP servers and split work among sub-agents. You pay according to the prices of the model, the tools and the execution environment.

The same day, Computer use was added to the Agents API as well: an agent that works in a browser hosted by OpenAI. Here is a detail worth noticing. According to the changelog, the developer's application is what approves which sites the agent can reach, and it is responsible for logging in to accounts. Again, the hands on the wheel.

What does this mean for a business? Directly, almost nothing. Indirectly, a lot. Whoever builds agents for businesses gets steadier infrastructure, so less time goes into "plumbing" and more into fitting the agent to the business itself. If you're curious what this looks like from the inside, we wrote about building an AI agent.

And what else: Codex, Space, sign-in with ChatGPT and privacy

  • Codex Cloud. Coding tasks run in the cloud and keep working even when your computer is asleep. You can check the results from the web, your phone or your computer.
  • ChatGPT Space and Pages. According to BGR, a shared workspace for team members and their dots, and a document editor where people and agents work together.
  • Sign in with ChatGPT. Developers can let users sign in with a ChatGPT account and use their subscription quota in the app. According to BGR, at launch this covers 16 partners, including Notion and Vercel.
  • MCP Events. A connected app can send ChatGPT an update when something happens, and the user defines in advance what to do with it. For example: when a new comment arrives on a document, make the fix.
  • Privacy for organizations. OpenAI showed a preview of OpenAI Private Intelligence, per Engadget. The documentation already lists ZDR with Private Safety Processing: content is stored encrypted in the customer's own storage, and safety checks run in a closed environment with no human access.
  • Plans. According to Engadget, a Pro 500 tier was added. According to BGR, the Pro 200 tier has reopened to new subscribers, but with a different quota than existing subscribers have.

What's relevant to a business now, what comes later, and what's only for developers

DevDay 2026 announcements by relevance to a small business

Worth knowing now

  • dots, if you have Pro or Business Premium
  • GPT-6.1 Sol, through whoever builds your tools
  • ChatGPT Space, if you work in a team

Watch over the coming months

  • dots expanding to more plans and markets
  • Sign in with ChatGPT in the apps you use
  • Teams of dots

Mainly for developers

  • Agents API and Computer use
  • Ultrafast
  • MCP Events
  • ZDR with Private Safety Processing

One conference, four points of view

Pick who you are, and see what's worth taking from DevDay 2026.

Small business owners

So what do I need to do tomorrow morning?

Probably nothing urgent. Three things to keep in mind:

  • Agents that work in the background have become an off-the-shelf product. This is the time to think about which tasks you would hand over.
  • Cheaper models at a high level can make long tasks worthwhile.
  • The rules matter as much as the capabilities. In any tool, check what it does on its own and what needs your approval.

ChatGPT users

What will change for me in the app?

If you're on a Pro or Business Premium plan, you may soon see an option to create a dot. First-time setup is done in the desktop app, and then you can continue from your phone.

✓ First dot included · ! Tasks count toward the quota

Teams and organizations

Is there anything here for our team?

ChatGPT Space and Pages are meant for people and agents to work together on the same files. In Enterprise, the system administrator turns dots on. And if privacy matters to you, keep an eye on OpenAI Private Intelligence.

Developers

What should I check this week?

  • Run your tasks on GPT-6.1 Sol and compare against Astra
  • Try the Agents API, in beta, with Computer use
  • Check whether Ultrafast justifies a price 6 times higher
  • Support MCP Events in your plugin

Our own summary and interpretation based on the announcements, not a quote from OpenAI.

Hype versus reality: what we think

What you heard at the conference

The headlines

  • An agent that works for you 24/7
  • Almost Astra, at one fifth of the price
  • Up to 8 times faster
  • More than 4,000 connected apps

What's worth checking

The bottom line

  • Which actions it takes without asking
  • How it handles your tasks, not benchmark tests
  • Whether the speed is worth a price 6 times higher
  • Whether your systems are really on the list, and with what permission

Our opinion: the decision not to release GPT-6.1 Astra is the most important news of the week, even though it wasn't announced from the stage. According to an OpenAI statement published by CBS News, the model fell short of the bar on staying within the limits of its task and permissions, and on reporting to the user what it actually did. That is exactly what worries a business owner who is considering giving an agent access to the website or the ad account.

The conclusion is simple: agent capability is rising fast, so the limits have to be defined in advance, not after something goes wrong. Every publication, budget change or sensitive action should wait for your approval, no matter which model the agent runs on.

What this means for Platin Digital clients

Platin Digital is an OpenAI Select Partner, and the agents we set up are built with whichever models and tools fit each business best, from OpenAI, Anthropic, Google and others. They run on a separate cloud server, connect to WordPress, Google Ads, Meta and Analytics, and receive tasks on Telegram.

  • We don't move an agent to a new model on announcement day. First we test GPT-6.1 Sol on real tasks, check quality, cost and behavior, and only then decide whether and when to switch.
  • The approval rules don't change. Publishing, budget, deletion and sending a message on behalf of the business continue to wait for your approval, on every model.
  • We'll evaluate dots and the Agents API. If they fit some of the tasks, we'll say so. There are businesses for which a dot inside ChatGPT will be enough, and that's perfectly fine.

Frequently asked questions

What is OpenAI DevDay 2026?

OpenAI DevDay 2026 is OpenAI's annual developer conference, held on September 29, 2026 in San Francisco. More than 20 new releases were announced there, including dots, standing agents inside ChatGPT, the GPT-6.1 Sol model, the Ultrafast speed tier and the Agents API for developers.

What is the difference between dots and an AI agent set up for a business?

A dot is an agent inside ChatGPT that connects to apps through ChatGPT plugins and operates according to the rules you set in it. An agent set up for a business usually runs in a separate environment, connected directly to the business's systems with dedicated users and permissions, and tailored to its processes. Sometimes a dot will be enough, and sometimes you need a deeper connection. Our guide to an AI agent for your business goes into detail.

Are dots available in Israel?

According to reports, dots are rolling out gradually to Pro, Business Premium and Enterprise plans in select markets. OpenAI didn't publish a list of countries in the sources we checked. The safe way to find out is to check your ChatGPT account.

How much does GPT-6.1 Sol cost?

According to OpenAI's pricing page, $2 per million input tokens and $10 per million output tokens, on the standard tier and for requests of up to 272 thousand input tokens. That's one fifth of the price of GPT-6 Astra.

Why didn't OpenAI release GPT-6.1 Astra?

According to an OpenAI statement published by CBS News the day before the conference, the model didn't meet the company's safety bar: it tended to step outside the limits of its task and permissions, and didn't report accurately on what it had done. According to Engadget, OpenAI will investigate the root cause.

Will Platin's agents move to the new models automatically?

No. We test every new model on real tasks before moving an agent to it, and decide based on quality, cost and behavior. The approval rules stay the same in any case.

Sources

Chat on WhatsApp