OpenAI DevDay 2026: Dots, Codex Cloud and 20+ Confirmed Launches

OpenAI DevDay 2026 is over. The keynote took place on September 29 at Fort Mason in San Francisco, and OpenAI used it to introduce more than 20 products, features and developer tools in one sitting.

The headline announcement is Dots, an AI agent that is meant to keep working after you stop chatting with it. Instead of waiting for your next message, a dot gets its own cloud computer and browser and works toward a goal you set, then checks back in. Alongside it, OpenAI shipped Codex Cloud for coding tasks that keep running in the background, a shared team workspace called ChatGPT Space, a document editor called Pages, and an OpenAI Marketplace that lets enterprise customers put their OpenAI spending toward partner software.

Here is what OpenAI actually confirmed, what it means in plain terms, and how it’s different from what came out earlier in September.

What OpenAI confirmed at DevDay 2026

OpenAI framed the keynote as a look at where it wants ChatGPT and its developer tools to go next: agents that don’t stop working when you close the app, coding that runs in the cloud instead of on your laptop, and more ways for teams to work alongside AI rather than just prompting it one message at a time.

The company said it made more than 20 announcements on the day. Not everything on that list is a new product. Some are pricing changes, some are features rolling into tools that already existed, and a few are enterprise-only previews. This article covers the announcements that actually matter for developers and everyday ChatGPT users, and is clear about which is which.

Dots: OpenAI’s always-on AI agent

Dots is OpenAI’s new agent inside ChatGPT, and it was the biggest reveal of the day. The idea is simple to explain even if the engineering behind it isn’t: instead of a chat that ends when you stop typing, a dot keeps working on whatever you asked it to do, in the background, until it’s finished or needs your input.

Each dot runs on OpenAI’s GPT-6 Astra model and gets its own cloud computer and browser, separate from your own device. That means it can open apps, browse the web, and carry out multi-step tasks without needing your laptop to stay on. OpenAI says a dot can connect to workplace tools such as Slack and Microsoft Teams, and that it learns your preferences over repeated use. Text-message support is described as coming soon rather than available at launch.

OpenAI has been careful to describe this as agents working toward goals you set, with identities, credentials and tools scoped to what they need, rather than an assistant with unrestricted access to your accounts. The company hasn’t published a full breakdown of every approval step a dot requires before taking an action, so treat the control boundaries as a work in progress rather than fully documented.

Availability: Dots launched on September 29 for ChatGPT Pro 200 and Business Premium subscribers. Enterprise, Education and Healthcare customers can get access with admin approval. It is not available on Free, Plus or standard Pro plans at launch. The first dot is included in the subscription price, and OpenAI says additional dot conversations don’t count against your regular usage limits.

For background on the model powering Dots, see our earlier coverage of GPT-6 Astra and the safety concerns OpenAI raised about it before DevDay.

GPT-6 Sol and Luna were not launched at DevDay

One thing worth getting right: GPT-6 Sol and GPT-6 Luna are not DevDay announcements. OpenAI released both models on September 22, a full week before the keynote, as cheaper, faster follow-ups to GPT-6 Astra. GPT-6 Sol dropped its API price to $2 per million input tokens and $10 per million output tokens, roughly half of what its predecessor cost, while GPT-6 Luna became the low-cost option at $0.10 per million input tokens and $0.50 per million output tokens.

What OpenAI did confirm on September 29 is GPT-6.1 Sol, an upgraded version of Sol built specifically for agentic coding, computer-use tasks and professional work in ChatGPT Work and Codex. OpenAI says it’s priced at roughly a fifth of what Astra costs per token, sitting between standard GPT-6 Sol and Astra in capability. It is fair to say OpenAI expanded GPT-6.1 Sol in Work and Codex at DevDay. It is not accurate to say OpenAI launched GPT-6 Sol or Luna at DevDay, since both were already available a week earlier.

Codex Cloud lets coding tasks keep running

Codex, OpenAI’s coding agent, now has a cloud mode. Codex Cloud runs your coding tasks in a hosted, isolated environment, so work can continue after you close your laptop, and you can start or check on a task from your phone. Each task gets its own environment with your repositories, tools and dependencies configured, rather than sharing a single sandbox with everything else you’re working on.

OpenAI also updated the Codex command-line tool with two-way voice input, a new view for tracking agent activity, and the ability to resume a session where you left off. Codex Cloud is available now on Plus, Pro, Business, Healthcare, Education and Enterprise plans; it is not a Free-tier feature.

ChatGPT Space and Pages bring teams together

ChatGPT Space is a shared, persistent workspace for a team and the dots working on their behalf. OpenAI describes it as a place where a team, ChatGPT and its agents can keep shared context and work on a project together, pulling in the kind of information you’d normally split across Slack, Notion or Google Drive, with AI tools built in rather than bolted on.

Pages is a new document type meant for people and agents to edit side by side. It supports writing, research, chart generation and image creation inside the same document, so a dot can draft or update a section while a human edits another part of the same page. OpenAI has said collaborative slides are coming within weeks of DevDay, but that feature had not shipped as of the keynote.

Connected tools and Sign in with ChatGPT

Alongside Space and Pages, OpenAI expanded how teams manage the tools their agents can use. Admins can approve which connected apps a dot or Codex agent is allowed to reach, rather than granting blanket access. OpenAI also introduced Sign in with ChatGPT, a single sign-on system that lets Plus and Pro subscribers use their existing OpenAI usage allowance across 16 partner platforms, including Notion, Vercel and Devin, instead of paying for a separate subscription on each one.

OpenAI Marketplace expands enterprise software access

For enterprise customers specifically, OpenAI introduced the OpenAI Marketplace. Eligible enterprise customers can apply part of their existing OpenAI spending commitment toward approved partner software instead of paying for it separately. OpenAI named an initial group of around 32 partners spanning creative tools (Adobe, Figma), customer experience (Sierra, Decagon, HubSpot, Salesforce, ServiceNow), legal (Harvey, Legora), security (Palo Alto Networks, CrowdStrike) and open-source model hosting (Baseten). This is an enterprise-only program; it does not change what individual ChatGPT subscribers pay or get.

Other confirmed DevDay announcements

• Ultrafast tier — a paid speed option that OpenAI says is up to eight times faster in Codex and up to six times faster through the API, currently running on GPT-6 Astra for Pro 500 and Enterprise plans, with a faster GPT-6.1 Sol version to follow.
• Agents API — now in public beta, giving developers hosted agent execution, memory and multi-agent tooling without building that infrastructure themselves.
• Codex Code Review and Security Cloud — automatic diff summaries and change analysis, plus scheduled vulnerability scanning for repositories connected to Codex.
• Pricing changes — the Pro plan’s usage multiplier for new subscribers dropped from 20x to 10x Plus usage (existing Pro subscribers keep 20x), while Plus usage limits increased; the Pro 200 tier reopened to new subscribers after a pause that began September 10.

What this means for everyday users

For most people, the practical change is less repeating yourself. Today, a lot of ChatGPT use involves giving an instruction, waiting for a reply, then giving the next instruction. Dots is built around the opposite pattern: you describe an outcome, and the agent keeps working toward it across a longer stretch of time, checking in when it needs a decision from you.

That’s only available on higher-tier plans for now, so most Free and Plus users won’t see it immediately. What will reach more people sooner is Codex Cloud, if you write code, and Sign in with ChatGPT, if you use any of the 16 partner apps OpenAI named.

What this means for developers

Developers get a heavier set of changes. The Agents API moving to public beta means hosted agent execution, memory and multi-agent orchestration are available without standing up that infrastructure separately. Codex Cloud and its new CLI features change where and how coding tasks run, not just what model is behind them. And GPT-6.1 Sol gives a cheaper, agent-tuned option for coding and computer-use workloads that sits below Astra’s cost. Teams building on connected apps should also check OpenAI’s new MCP event support, which lets connected apps trigger automations rather than only responding to direct requests.

Frequently asked questions

Sources

• OpenAI: DevDay 2026 Recap
• TechCrunch: OpenAI launches Dots, its always-on agentic assistant
• OpenAI Developer Community: DevDay 2026 announcements and developer resources
• OpenAI: Introducing GPT-6 Sol and Luna

Related articles

Comments