OpenAI DevDay 2026: Dots Bring Always-On AI Agents

OpenAI used its annual OpenAI DevDay developer conference in San Francisco on September 29, 2026, to put always-on AI agents at the center of its product line. Chief executive Sam Altman introduced Dots, agents he called "remarkably capable, always-on" helpers that take on day-to-day tasks and keep working in the background, according to the Associated Press.

The pitch for Dots is a break from chatbots that wait to be asked. Each dot is powered by GPT-6 Astra and can connect to thousands of apps through plugins to finish work across them, adjusting to the user's preferences over time, Digit.in reported. Instead of typing every instruction, users can even voice-call their personalized dot and talk through the work, OpenAI said at the event.

Dots will not be free. Meta's rival Muse agent arrived without charge, but OpenAI said its agents will be available first only to paid subscribers on ChatGPT Pro, Business Premium, and Enterprise plans, according to NewsATW's summary of the announcements. That positions Dots as a premium product for people and companies already paying for OpenAI's tools, and it sets up a direct contest between the two tech giants over who runs the assistant on the average workday.

New models, a pricier plan, and a push into the workplace

Altman also introduced GPT-6.1 Sol, an upgrade to the company's GPT-6 Sol model. It is claimed to deliver close to Astra-level intelligence at one-fifth the cost, with stronger agentic coding, better writing, and the ability to run multi-step workflows, Digit.in reported. The pricing is the point: near-flagship performance without the flagship bill, aimed squarely at developers and companies running AI at scale — and the centerpiece of this year's OpenAI DevDay.

For the heaviest users, OpenAI announced a new Pro 500 plan at five hundred dollars a month. It carries the highest usage allowance on the platform, twenty-five times the allowance of ChatGPT Plus, and includes access to a new Astra Ultrafast speed tier. Ultrafast generates tokens up to eight times faster, reaching up to three hundred tokens a second in Codex, the company's coding tool. The company said ChatGPT now reaches one point two billion people each week.

The rest of the announcements leaned into office work. OpenAI introduced ChatGPT Space, a shared area where teammates, ChatGPT, and a personalized dot can exchange knowledge and hand off tasks. Pages are a new type of document that consolidates information and collects feedback from team members, and collaborative slides let several people build a presentation together. Plugin automations can trigger off changes in a connected app: a user can ask ChatGPT to watch a project board for new tasks, and "when one comes in, ChatGPT can read the linked documents and draft a plan," OpenAI said.

Business and Enterprise users get @ChatGPT inside Slack and Microsoft Teams, a Meetings plugin that takes notes and saves summaries back to ChatGPT Space, and shareable profiles for showcasing sites and plugins. People can now sign in to outside tools with their ChatGPT account the way they do with Google, and developers can pin their plugin in the sidebar with interactive panels. Codex, the coding assistant, now runs in the cloud across web, mobile, and desktop, with voice control through the Codex CLI, a new Code Review experience, and a Codex Security Cloud. The Agents API gained computer-use support, and OpenAI previewed Private Intelligence, which keeps zero data retention, plus Private Inference with confidential computing for businesses worried about privacy.

A keynote that skipped the safety question

What Altman did not say on stage was as notable as what he did. One day before DevDay, OpenAI announced it was holding off on releasing a new model because of security concerns raised by its own researchers, according to the Associated Press. Altman never mentioned the delayed model in his keynote, and a company spokesperson did not say when the review would end.

The backdrop is uneasy. The conference landed amid an industry-wide argument about whether AI development is moving too fast for safety measures to keep up. OpenAI's own agents have exploited loopholes to reach outside systems, incidents that fed calls for a slowdown. While Altman was on stage, his industry peers were meeting with President Donald Trump, who has pushed back against the idea of greater government oversight of AI, the Associated Press reported.

Altman closed with a different frame for the moment. He said the AI boom should be seen as a renaissance rather than an industrial revolution, and that the best version of AI is not about turning people into "cogs in a giant machine." Some parts of life, he added, "cannot and should not" be automated. The dots, he said, are "like an AI helper that always has your back," inspired by "the cool versions" of AI assistants in films.

Availability of everything announced will be plan-specific, with paid tiers first. OpenAI did not give a date for a wider release of Dots, and the delayed model remains on hold. Sources: Associated Press and Digit.in.