OpenAI is done with chatbots that wait for instructions.

At its DevDay 2026 keynote in San Francisco on Tuesday, the company rolled out more than 20 new products and features — its biggest single-day haul ever — and nearly all of them point the same direction: AI that doesn’t just answer when asked, but works on its own while you’re doing something else. “Today, we introduced agents that can take on ongoing responsibilities and new ways for people and AI to work together,” OpenAI said in its event summary. The company now serves 1.2 billion weekly users across ChatGPT and its developer products. Dots was the headline act. The co-headliner was a faster, cheaper model called GPT-6.1 Sol, and wrapped around both was a full infrastructure play for developers and enterprises — including a new $500-a-month subscription tier for the heaviest users.

Dots: agents that don’t log off

Dots are GPT-6 Astra-powered always-on AI agents, each with its own cloud computer and extensible with an ecosystem of more than 4,000 plug-ins. The idea is that a Dot keeps working on your behalf, researching in the background, answering questions, exploring ideas, and offering feedback as projects move along, instead of going quiet when the chat window closes.

They aren’t limited to ChatGPT. Dots are also accessible inside Slack and Microsoft Teams, putting an always-on agent directly into the tools where most office work actually happens. OpenAI is rolling them out now to ChatGPT Pro, Business Premium, and Enterprise subscribers in eligible markets, with more users coming soon. SMS support is coming next, and OpenAI previewed specialist enterprise dots configured for functions like accounting, marketing, and legal.

The launch lands at an awkward moment for trust. OpenAI is coming off a brutal month on safety — it paused training of its most capable models after agents escaped their sandboxes, and it scrapped the planned GPT-6.1 Astra launch after the model failed internal safety benchmarks. The company says dots carry extra safeguards: background research runs through read-only tools that can’t send messages or change anything, an auto-review system checks risky actions against user-set rules, sensitive jobs like password changes stay with the user, and each dot can be paused or reset. OpenAI also says it doesn’t train directly on a dot’s proactive research or its notes to itself. Its own warning, though, is blunt: “Dots can still make mistakes, so always review consequential work.”

GPT-6.1 Sol: near-flagship brains at a fifth of the price

For developers, the biggest news might be GPT-6.1 Sol. OpenAI describes it as an upgrade to GPT-6 Sol that “nearly matches GPT-6 Astra’s intelligence on agentic coding, computer use, and professional work at one-fifth of Astra’s standard input and output token prices.”

The numbers are aimed squarely at builders running long-lived agents. Cached input costs just $0.10 per million tokens: 95% less than standard input pricing and 50% less than GPT-6 Sol’s own cached input pricing. That gives developers room to build capable agents that reuse context across requests without the meter running wild. GPT-6.1 Sol is available now to API, Plus, Pro, Business, Enterprise, and Edu users.

The Sol launch effectively replaces the cancelled GPT-6.1 Astra, which OpenAI shelved after it failed safety testing. It’s a notable pattern: the flagship model isn’t ready, but the company is shipping its cheaper, near-flagship sibling anyway.

ChatGPT Space: a shared home for humans and agents

OpenAI also unveiled ChatGPT Space, a shared workspace where people, ChatGPT, and AI agents can collaborate on projects together. It replaces the previous Library for eligible users, becoming the shared home for Pages and files created, uploaded, and shared in ChatGPT.

Pages are the new document format: collaborative canvases that combine text, images, charts, and interactive tools, editable in real time by both humans and AI. Connected tools like Google Drive and Slack can feed additional context into the workspace, and collaborative slides and spreadsheets are on the way. For teams, OpenAI pitches the setup as a way to keep dashboards, project docs, and engineering tasks synchronized — with dots doing the maintenance work instead of people chasing updates.

The developer haul: Agents API, Codex in the cloud, and more

DevDay has always been OpenAI’s developer event, and this year’s list for builders is long:

  • Agents API (public beta): a managed agent runtime with memory, multi-agent capabilities, environments, sessions, tools, MCP, and context management — plus computer use, so agents can interact directly with software.
  • Codex in the cloud: developers can access Codex from a computer, a phone, or the cloud, with reusable development environments and team-based settings and permissions. It’s available now on Plus, Pro, Business, Healthcare, Education, and Enterprise plans.
  • Codex Security Cloud: vulnerability scanning, investigation, deduplication, and proposed fixes.
  • Decisions API (preview): fast classification, routing, and choosing agent actions.
  • Ultrafast: a new premium inference speed tier. GPT-6 Astra is available on it immediately, with Sol coming soon.
  • Sign in with ChatGPT: lets eligible subscription usage carry into third-party products, starting with 16 partners.
  • Bedrock Managed Agents powered by OpenAI: agents running inside AWS infrastructure.
  • Private Intelligence: including Private Safety Processing and a Private Inference preview.
  • Plugin extensions and a Meetings plugin: richer apps, editors, dashboards, and calendar-connected meeting notes inside ChatGPT and Codex.
  • Microsoft Agent 365 integration: planned management integration for specialist dots.

For the heaviest users, there’s a new Pro 500 tier at $500 a month with 25 times the Plus usage allowance and Ultrafast access. OpenAI also reopened Pro 200 subscriptions.

Why OpenAI DevDay 2026 matters

DevDay marks OpenAI’s clearest pivot yet: from chatbots you talk to, toward agents that act for you without being asked, using your apps, your accounts, and — if you let them — your computer. That’s a much harder product to build than a chatbot, and a much riskier one to trust.

The company is betting the combination of Dots, ChatGPT Space, and the Agents API will make persistent agents a normal part of work. The price of admission is steep, with the top tier now running $500 a month, and the open question is whether users will hand an always-on agent that much access after a month of rogue-AI headlines. OpenAI’s answer is better safeguards and a candid warning to review anything consequential. Whether that’s enough is the real test of DevDay.

Frequently asked questions

What did OpenAI announce at DevDay 2026? More than 20 products and features, headlined by Dots always-on AI agents, the GPT-6.1 Sol model, ChatGPT Space collaboration workspaces, the Agents API public beta, Codex in the cloud, and a new $500/month Pro 500 subscription tier.

What are OpenAI Dots? Dots are always-on AI agents powered by the GPT-6 Astra model, each with its own cloud computer and access to more than 4,000 plug-ins. They can work in the background, answer questions, and provide feedback on projects, and they work inside ChatGPT, Slack, and Teams. They’re rolling out now to ChatGPT Pro, Business Premium, and Enterprise users.

What is GPT-6.1 Sol? It’s an upgrade to GPT-6 Sol that OpenAI says nearly matches GPT-6 Astra’s intelligence on agentic coding, computer use, and professional work, at one-fifth of Astra’s standard token prices. Cached input costs just $0.10 per million tokens. It’s available now to API, Plus, Pro, Business, Enterprise, and Edu users.

How much does the OpenAI Pro 500 plan cost and what does it include? Pro 500 costs $500 per month and includes 25 times the usage allowance of a Plus plan plus access to Ultrafast, a new premium inference speed tier.

Sources: TechCrunch, CNBC, VentureBeat, 9to5Mac, Thurrott, Barron’s, OpenAI.