OpenAI DevDay 2026: 6 Announcements That Matter for Small Businesses

The short answer

OpenAI DevDay 2026, held on September 29, 2026, was mostly about agents that keep working without you: always-on "dots" inside ChatGPT, computer use in the Agents API, and a cheaper GPT-6.1 Sol model. For a small business, the useful pieces are a lower price for capable automation, a faster way to route and classify requests, and new ways to show up inside ChatGPT. Most of it is still rolling out, so plan pilots, not overhauls.

Developer conferences are written for developers. This recap is written for the owner or manager who has to decide what is worth a test this quarter. Every product detail below comes from OpenAI's own documentation and help pages as of October 2026; where something was only reported by press or community posts, we say so.

What OpenAI DevDay 2026 announced, in one list

OpenAI described more than 20 announcements. A community-maintained index on OpenAI's developer forum links each one to its official page, and InfoQ's recap covers the developer side. The six that matter most to a small or mid-size business:

1. Dots: persistent, always-on agents in ChatGPT with their own cloud computer and browser.
2. GPT-6.1 Sol: a mid-priced model aimed at coding, computer use and professional work.
3. Computer use in the Agents API: agents that operate websites through a browser OpenAI hosts.
4. The Decisions API: fast, typed answers for classifying and routing things like emails and enquiries.
5. Plugin upgrades: richer plugin interfaces inside ChatGPT, plus "MCP Events" that let outside apps trigger work.
6. Team features: ChatGPT Space, shared Pages, and ChatGPT inside Slack and Microsoft Teams.

Dots: an assistant that keeps working after you close the tab

OpenAI's help documentation describes a dot as "an always-on agent that keeps work moving and asks for your input when a decision needs you." Each dot runs on GPT-6 Astra, lives in the cloud with its own computer and browser, and can keep working while your laptop is off, according to OpenAI's Meet dots page. It can follow up on tasks, run recurring check-ins, and be messaged from ChatGPT, Slack, Teams or by phone call.

Availability is narrower than the keynote suggested. As of October 2026, dots are rolling out gradually to Pro 100, Pro 200 and Pro 500 users over 18 outside the EEA, UK and Switzerland, and to Business Premium and Enterprise workspaces worldwide. In Enterprise, dots are off until an administrator turns them on.

Why it matters: this is the first mainstream product that behaves like a junior assistant rather than a chat window. It also raises new questions about what an agent may send, buy or delete on your behalf. We cover those in detail in our guide to always-on AI agents.

GPT-6.1 Sol: capable automation at a fifth of the flagship price

For anyone paying for API usage, pricing is the headline. On OpenAI's pricing page, standard short-context rates per million tokens are:

Model	Input	Output	Positioned for
GPT-6 Astra	$10.00	$50.00	Hardest reasoning and coding
GPT-6.1 Sol	$2.00	$10.00	Balance of intelligence and cost
GPT-6 Luna	$0.10	$0.50	High-volume, cost-sensitive work

InfoQ quotes OpenAI saying Sol "approaches GPT-6 Astra on several evaluations." Treat that as a vendor claim until you have tested it on your own documents. The practical takeaway: most business automations (summarising enquiries, drafting replies, extracting fields from invoices) do not need the flagship model. Testing Sol or Luna first can cut running costs sharply, which changes the maths in our breakdown of what AI automation really costs a small business.

Computer use and the Decisions API: the developer tools worth a pilot

Computer use in the Agents API

The Agents API can now run a browser in an OpenAI-hosted environment so an agent can navigate websites and use applications through their interface. The computer use guide is careful about limits: the browser asks the user to approve each new website origin, and OpenAI warns that origin approval "does not enforce confirmation before individual actions." If you need a guaranteed confirmation before a purchase or deletion, the guide says to restrict the browser or use a runtime you control. That is the right instinct, and we explain why in what computer-use agents can and can't reliably do.

The Decisions API

This one is quietly useful. According to the Decisions guide, the API returns typed answers about 10 times faster than the Responses API: a probability that a condition is true, a choice from a fixed list, or a score against a rubric. It currently supports only GPT-6 Luna and is in public beta (InfoQ described it as a limited preview at launch). Think "which department should this email go to?" or "is this enquiry urgent?" These narrow, repeatable calls are exactly where small businesses get dependable value from AI.

Plugins: new ways to appear inside ChatGPT

ChatGPT's former "apps" are now plugins, packaged with skills, an MCP server and optional interface. DevDay added sidebar panels, file viewers and MCP Events, which let a connected app trigger work when something changes. OpenAI's MCP Events documentation gives examples such as watching a Slack channel for bug reports. There is also a partner beta for local services in which ChatGPT shows a "Get Quote" button on eligible business listings. That is not open to everyone yet, but it signals where local search inside ChatGPT is heading.

If you are wondering whether your business should build one, read our guide to building a ChatGPT app or plugin first. For background on the protocol underneath, see our MCP glossary entry.

What to do this quarter

• If you are on an eligible plan: give a dot one low-risk recurring job, such as a weekly summary of unanswered enquiries, and set it to ask before sending anything.
• If you run API automations: re-test your prompts on GPT-6.1 Sol or Luna. Keep the flagship only where quality drops.
• If you sort lots of inbound messages: prototype routing with the Decisions API before building anything bigger.
• If customers search for services like yours: watch the plugin and Get Quote programmes, and keep your public business data accurate.
• Skip for now: anything that lets an agent move money or change passwords unsupervised. OpenAI's own docs keep those steps with a human.

Key takeaways

• OpenAI DevDay 2026 shifted the focus from chat to agents that run in the background.
• GPT-6.1 Sol costs one-fifth of GPT-6 Astra per token, which makes everyday automation cheaper.
• Computer use is available through the API, but per-action confirmation is your responsibility.
• Most launches are rolling out by plan and region as of October 2026, so check availability before planning around them.

Frequently asked questions

When was OpenAI DevDay 2026?

The keynote took place on September 29, 2026. Some coverage dates its recaps to early October because write-ups followed over the next few days.

Can a small business use dots on a Plus or Business plan?

Not as of October 2026. OpenAI lists Pro 100, Pro 200 and Pro 500 (outside the EEA, UK and Switzerland), Business Premium and Enterprise as eligible, with a gradual rollout.

Is GPT-6.1 Sol good enough to replace the flagship model?

For many routine tasks it may be, and it is far cheaper. OpenAI says it approaches Astra on several evaluations, but test it against your own examples before switching anything customer-facing.

Sources

1. DevDay 2026 announcements and developer resources — OpenAI Developer Community, September 29, 2026
2. OpenAI DevDay 2026 Recap for Developers — InfoQ, October 2, 2026
3. Meet dots — OpenAI (ChatGPT Learn)
4. Pricing — OpenAI API documentation
5. Computer use (Agents API) — OpenAI API documentation
6. Decisions — OpenAI API documentation
7. MCP Events — OpenAI Plugins documentation

How Flowmatix can help

We put this into practice for businesses in Toronto and around the world through our AI Agents & AI Solutions, business process automation, seo & performance optimization services. Want to know what it would take for your business? Get a personalised quote in about 3 minutes.