OpenAI DevDay 2026: Dots, Pro 500, and ChatGPT as a Billing Layer

OpenAI's DevDay 2026 on September 29 was not a single model drop. It was a stack of product bets aimed at people who already live in ChatGPT: always-on agents with their own cloud computers, a new $500 Pro tier, a quieter cut to Pro 200 usage, and a way for Plus and Pro subscribers to spend that same plan inside Notion, Devin, Vercel, and a dozen other tools.

The official recap counts more than 20 announcements. For marketers, sales teams, and anyone budgeting AI seats, five threads matter most: the Pro ladder, Dots, apps that draw on your ChatGPT plan, the new office-like collaboration surface, and a short list of developer launches that will show up in the tools you already buy.

Pro 200 is open again. The allowance is not what it was.

OpenAI reopened Pro 200 to new signups on stage. The price is still $200 a month. The included usage is not.

OpenAI's Pro help materials confirm that Pro 200 is open again with a lower included allowance for new subscribers than the plan carried before OpenAI paused new signups earlier in September. Existing subscribers keep their current limits through October 29, 2026, then move to that lower allowance. Press reporting based on OpenAI briefings and those Pro help materials, including The Next Web and Engadget, gives the size of the cut: from October 30, 2026, included ChatGPT Work and Codex usage falls from 20 times the Plus allowance to 10 times, and GPT-6 Pro messages in chat fall from 200 to 100 a week. The Next Web reports a one-time usage-credit package for existing Pro 200 subscribers, at $2,500 equivalent, that expires at year end. OpenAI also said it will not bring back Pro 200's five-hour limit.

Above that sits the new Pro 500 plan at $500 a month. OpenAI describes it as the highest Pro usage allowance, at 25 times ChatGPT Plus, and the only Pro plan that includes Ultrafast. Ultrafast is the paid speed tier for GPT-6 Astra: OpenAI says up to about 300 tokens per second in Codex, up to eight times standard speed there and up to six times in the API. Pro 100 remains on the ladder as the entry Pro seat.

For teams that treated Pro 200 as "pay once, run heavy agent work all month," the October 30 cut is the real news. The $500 tier is where OpenAI parked the speed and the bigger Work and Codex budget. Anyone grandfathered through late October should decide before then whether the old allowance was the product they bought, or whether Pro 500 (or a Business Premium seat) is the new baseline.

Dots: an always-on agent with its own computer

Dots are OpenAI's new personal agents, powered by GPT-6 Astra. OpenAI's launch post describes them as always-on agents with their own cloud computer, access to more than 4,000 apps through plugins, and the ability to keep working toward goals after you close the tab. They show up in ChatGPT, Slack, and Microsoft Teams, with texting described as coming later. OpenAI also sketched specialist dots with their own identity and credentials, including a path into Microsoft's Agent 365 admin controls.

The packaging for paid users is simple on paper. Your first dot is included with Pro and Business Premium in eligible markets, at no extra subscription fee. OpenAI has said conversations with your dot do not count toward ChatGPT usage limits. Work and Codex tasks the dot starts still consume the usual plan allowance. Enterprise, Edu, and Healthcare get a beta that admins can leave off by default.

Availability is regional. Official posts describe a gradual rollout for eligible Pro and Business Premium accounts, and they do not name those markets. It is reported that OpenAI told press Dots will not reach Pro users in the European Economic Area, Switzerland, or the UK at launch. That matters if your marketing org spans those markets: the agent story and the seat you buy may not line up yet.

For marketers, the useful comparison is not "another chatbot." It is closer to the Grok Bot idea of a teammate with a machine, or Block's Buzz agents on the org chart. Dots put OpenAI's version of that pitch inside ChatGPT, with plugin reach as the distribution layer. Treat consequential sends, spend, and CRM writes the way OpenAI's own safety framing suggests: review the work, and keep sensitive actions behind confirmation.

Your ChatGPT plan can pay for work inside other apps

Sign in with ChatGPT now goes past identity. Plus and Pro users can authorize eligible OpenAI usage inside partner apps so those requests draw on the ChatGPT Work and Codex allowance already in the plan. You do not paste an API key. The app still may charge for its own product.

OpenAI's DevDay recap confirms 16 partner apps that can draw on plan usage, and it names six of them: Devin, Notion, Vercel, Dactyl, OpenClaw, and T3. Secondary coverage, including NerdsChalk and RuntimeWire summarizing OpenAI's list, fills out the rest of the 16. That reporting adds the commercial tools Amp Code, Conductor, Hermes Agent, Hyperagent, Kilo Code, Lovable (coming soon), Vorflux, and Warp, and the open-source tools OpenCode and Pi.

A separate group, including Airtable, Canva, GitLab, HubSpot, and Supabase, supports ChatGPT sign-in only. Logging in there does not spend your plan. Plan sharing is a separate permission. OpenAI says connecting an app does not by itself hand over ChatGPT conversations or memories; sign-in shares name, email, and profile picture.

Caps live in ChatGPT under Settings → Usage. Each connected app gets a weekly limit as a percentage of your overall weekly usage. That cap does not reserve capacity for the app, and it does not raise your total. After you hit plan limits, an app can keep going on ChatGPT credits only if you turn credits on for apps and set that app's cap to 100%. Credits and automatic top-ups are off by default for a reason: a busy Devin or Warp session can burn through a week's Work and Codex budget faster than a normal ChatGPT day.

For marketing ops, the strategic read is that ChatGPT just became a billing and distribution layer, not only a destination. Notion and Vercel on that list matter as much as the coding agents. Budget the seat as shared capacity across tools, not as a single product line item.

ChatGPT wants to be the shared office

DevDay also pushed ChatGPT further into team collaboration. ChatGPT Space replaces Library as a shared home for a team's files and documents. Pages are the main format: documents teammates, ChatGPT, and dots can edit together, including updates from connected tools. Space is rolling to Pro, Business, and Enterprise on web and desktop, with mobile read and share first and editing to follow.

Collaborative slides are due in the coming weeks on the same plans, with export to PowerPoint or Google Slides. Business and Enterprise can form teams inside ChatGPT and set Team Tasks that run on a schedule or when something happens in a connected channel. People can @chatgpt in Slack and Microsoft Teams channels, threads, and DMs; teammates do not need their own ChatGPT licence for that mention flow, and OpenAI says it asks for approval before acting in connected tools.

A Meetings plugin, in beta on the Mac desktop app for Pro and Business, takes notes without a bot joining the call. OpenAI says audio is deleted once the notes are ready. Users are told to get consent from everyone in the meeting; the in-app reminder is visible only to the person running the plugin.

Taken together, Space, Pages, slides, shared tasks, and @chatgpt are OpenAI competing with Google Docs, Notion, and meeting note apps from inside the assistant people already open all day. The Sign in with ChatGPT Notion partnership makes that rivalry more interesting, not less.

What else shipped, briefly

A few other launches will show up in vendor roadmaps even if you never open the API console:

• GPT-6.1 Sol is OpenAI's mid-tier model pitch: near-Astra intelligence at list prices reported as $2 per million input tokens and $10 per million output, with cached input at $0.10. At launch it is aimed at API, Work, and Codex paths rather than ordinary Chat.
• Ultrafast pairs with Pro 500 and Enterprise for Astra speed; Sol Ultrafast is listed as coming soon.
• Codex gained cloud task running, a refreshed CLI with voice steering, Code Review from the desktop app into GitHub and GitLab, and Codex Security Cloud for repo scans and prepared fixes on higher plans.
• Decisions API (limited preview) focuses Luna on closed, user-defined answer sets for routing and classification.
• Agents API now includes computer use, with OpenAI hosting more of the agent loop.
• Bedrock Managed Agents put OpenAI-powered agents inside AWS for customers who need that boundary.
• Private Intelligence covers automated safety review without OpenAI staff seeing underlying content, with Private Inference preview due later.
• Plugin extensions, Plugin Creator, MCP Events support, and an OpenAI Marketplace for eligible US enterprises to spend part of OpenAI commitments on partner software (Adobe, Figma, CrowdStrike, and others in the beta set) round out the platform side.

What to do with it this week

If you run Pro 200 today, open Settings → Usage and decide before October 29 whether the lower Work and Codex allowance still covers your agent load. If you need Ultrafast or the 25× Plus budget, Pro 500 is the product OpenAI designed for that. If your seat is Pro or Business Premium in an eligible market, try one Dot on a bounded workflow with a human approval at the end: a research brief, a CRM cleanup draft, a competitive monitor. Do not hand it production sends on day one.

If your stack already includes Notion, Vercel, Warp, or Devin, turn on Sign in with ChatGPT plan usage only after you set per-app caps. Treat the ChatGPT seat as shared capacity across those tools. And if collaboration is the pain, Space and Pages are the features to watch against the docs and decks your team already trusts.

DevDay 2026 made the direction clear. ChatGPT is no longer only the place you ask questions. It is becoming the place work keeps running, the place teammates co-edit with agents, and the plan that other apps bill against. The $500 tier and the Pro 200 cut are how OpenAI is pricing that reality.