OpenAI DevDay 2026: Every launch, price and access rule
OpenAI announced more than 20 launches, but access depends on your plan, region and administrator. Check the prices, previews and billing rules before you subscribe or build.
OpenAI DevDay 2026 brought a new $500 Pro plan alongside a cut to the $200 plan's future allowance. You can't use every launch today: some are previews, some need an administrator, and collaborative slides are still coming.
OpenAI held DevDay on 29 September in San Francisco. Its official recap lists more than 20 announcements, checked below against product documentation as of 30 September. Before you subscribe or build, check access and billing.
OpenAI DevDay 2026: every launch, with prices and access
The table draws on the launch recap, DevDay documentation and TNW's launch coverage. "No separate price stated" means no published launch price. It doesn't mean unlimited use or free API calls.
| Group | Announcement | What changed | Access and price |
|---|---|---|---|
| Agents | dots | An ongoing agent with its own cloud computer and connected apps. | First dot included with eligible Pro or Business Premium subscriptions. Regional and admin restrictions apply. |
| Agents | Specialist dots | Business agents with their own identities and credentials, including Microsoft Agent 365 integration. | Enterprise-focused pilots; general rollout and price not stated. |
| Models | GPT-6.1 Sol | Updated coding and work model, available as gpt-6.1-sol. |
API; Work and Codex on Plus, Pro, Business, Enterprise and Edu. Standard short-context rates: $2 input and $10 output per million tokens. |
| Models | Ultrafast | Faster token generation for GPT-6 Astra. | API; Work and Codex on Pro 500 and eligible Enterprise and Edu plans. Sol support is coming. |
| Privacy | Private Intelligence | Private Safety Processing supports automated safety checks without staff seeing underlying content. | Business offering; no public launch price. Private Inference preview is due this fall. |
| Codex | Cloud tasks | Remote coding work continues while your computer sleeps. | Plus, Pro, Business, Healthcare, Education and Enterprise; uses plan allowance. |
| Codex | Reusable cloud environments | Prepared development setups for new isolated tasks. | Part of the cloud workflow, with approved settings and permissions. |
| Codex | Refreshed CLI | Voice steering, an /agents view and session improvements. |
All plans, subject to their usage limits. |
| Codex | Code Review | Desktop summaries and diffs, GitHub and GitLab feedback, automatic cloud reviews. | All plans, subject to usage limits. |
| Codex | Codex Security Cloud | Repository scans, new-commit monitoring, validated findings and proposed patches. | Research preview on Pro, Business, Enterprise and Edu; includes Daybreak Blue model access. |
| API | Decisions API | GPT-6 Luna picks from developer-defined answers using text or image context. | Limited preview; broader release planned in the coming days. No separate price stated. |
| API | Agents API with computer use | Hosted agents with tool calling, tool search, multiple agents and context compaction. | API; Work and Codex on Pro 500 and Enterprise. Check API billing separately. |
| API | Bedrock Managed Agents | OpenAI-powered agents running inside AWS. | AWS preview; no launch price stated on the product page. |
| Platform | Plugin extensions | Sidebar apps, conversation panels, file viewers and editors. | Broad plan support; Free and Go web extensions are coming soon. |
| Platform | Plugin Creator | Help building plugins, revised submission feedback and better discovery. | All plans in the recap; no separate launch price stated. |
| Platform | ChatGPT Sites with plugins | Workspace apps use each visitor's connected accounts and permissions. | Business, Enterprise, Healthcare and Edu; enabled workspace-private Sites. |
| Platform | Sites databases | SQLite storage for Sites, reported in the Sites session. | Session coverage does not specify a separate price or plan restriction. |
| Platform | Sites scheduled tasks | Sites can run scheduled work, reported in the Sites session. | Announced there as launching that day; separate price not stated. |
| Platform | MCP Events | Connected-app events can trigger automations. | Draft specification support, requiring MCP 2.0; no separate price stated. |
| Platform | Website annotations | Developers control selectable context and change previews in supported browsers. | Browser Annotation API in supported desktop app browsers; no separate price stated. |
| ChatGPT | Work across devices | Continue a new synced task on desktop, web or mobile. | Local-file steps still need an online, connected computer. Existing tasks retain their mode. |
| ChatGPT | ChatGPT Space | A shared home for files and Pages. | Pro, Business and Enterprise on desktop and web; mobile can find, read and share. |
| ChatGPT | Pages | Documents people and agents can edit together. | Pro, Business and Enterprise. |
| ChatGPT | Collaborative slides | Shared deck editing and PowerPoint or Google Slides export. | Pro, Business and Enterprise, coming in the next few weeks. |
| ChatGPT | Teams and Team Tasks | Shared work runs on a schedule or an event. | Business and Enterprise, through approved team connections. |
| ChatGPT | @ChatGPT in Slack and Teams | Ask ChatGPT inside channels, threads and direct messages. | Business and Enterprise deployments; participating teammates need no individual ChatGPT licence. |
| ChatGPT | Meetings plugin | Meeting notes without a bot joining the call. | Mac desktop beta for Pro and Business; Enterprise is coming. Audio is deleted after notes are ready. |
| ChatGPT | Shareable profiles | Show Sites and activity through a profile. | Free, Go, Plus, Pro, Business and Enterprise; personal profiles are private by default. |
| Enterprise | Workspace connections | Company-managed accounts support team workflows. | Managed permissions; no separate launch price stated. |
| Enterprise | Agent Security | Admin controls for tools, files, networks and execution. | Workspace policies; Work Cloud has separate controls. |
| Plans | Pro 500 | A higher allowance and Ultrafast access. | $500 monthly, with 25 times the Plus Work and Codex allowance. |
| Plans | Pro 200 | Reopened subscriptions with a lower ongoing allowance. | $200 monthly; eligible existing users retain old limits through 29 October. |
| Distribution | Sign in with ChatGPT | Eligible app requests draw on the user's ChatGPT plan. | Plus and Pro plan usage; 16 launch partners, selected commercial trial. |
| Procurement | OpenAI Marketplace | Eligible US enterprises apply part of an OpenAI spending commitment to approved software. | Beta with 32 partners; not a general consumer app shop. |
dots, Space and Pages come with different limits
A dot can keep pursuing a goal, Space holds its documents, and Pages give you somewhere to edit the result beyond an isolated chat. Permissions deserve as much attention as the model. Connecting an account shouldn't let an agent act everywhere.
The dots documentation includes Pro 100, Pro 200 and Pro 500 for adults outside the European Economic Area, UK and Switzerland. Business Premium rolls out worldwide; Enterprise access is worldwide but off by default. The recap also lists admin-enabled Edu and Healthcare beta access.
Conversations with your dot don't consume ChatGPT conversation allowance, but the Work and Codex tasks it starts still draw from the relevant pool. "First dot included" covers the dot in your subscription. It doesn't make every task the dot starts unmetered.
You can read and share Space from mobile; mobile creation and editing are still coming. Slides also have an expected release window, rather than availability everywhere today.
Ultrafast spends your allowance faster too
The API pricing page lists these standard rates for requests with no more than 272,000 input tokens. Prices are dollars per million tokens, before separate tool charges.
| Model or speed tier | Input | Cached input | Output |
|---|---|---|---|
| GPT-6 Astra, Standard | $10 | $1 | $50 |
| GPT-6.1 Sol, Standard | $2 | $0.10 | $10 |
| GPT-6 Luna, Standard | $0.10 | $0.01 | $0.50 |
| GPT-6 Astra, Ultrafast | $60 | $6 | $300 |
Long-context requests have different rates, so this table won't price every request you send. Luna's regular token price also shouldn't be read as a published Decisions API price.
OpenAI describes Sol as offering near-Astra performance at a fifth of Astra's standard input and output prices. The table lets you check the prices. Performance remains OpenAI's assessment, rather than an independent result established here.
On Hacker News, minimaxir's comment picked out cached input costing half the previous Sol rate. If you keep sending the same instructions and tool definitions, that cheaper cache deserves a closer look.
How you pay changes Ultrafast's consumption rate. The Work and Codex pricing guide says Astra Ultrafast uses included subscription allowance at eight times Standard's rate; purchased credits and enterprise usage use six times the Standard rate. Those multipliers describe consumption separately from the advertised speed increase, so faster generation can empty your allowance faster too.
Before you build, check where the work runs
In the Sites session after the keynote, Simon Willison reported SQLite databases and scheduled tasks launching for Sites. The table attributes them to that session reporting and keeps its limits explicit.
A cloud task can keep going with your laptop closed. A synced task that needs your laptop's files still needs that machine online. If you want unattended work, check where the task runs before assuming that switching screens lets you shut the laptop.
Security Cloud prepares evidence and patches for review before you create a draft pull request, as its setup guide describes. A scan helps your security process. It can't prove a repository has no vulnerabilities.
AWS calls Bedrock Managed Agents a preview. Keeping runtime and inference inside AWS is the attraction if you're already an AWS customer. Preview access still doesn't mean every account can immediately deploy every configuration.
Sign in with ChatGPT doesn't offer unrestricted commercial access. Marketplace handles procurement for eligible enterprises, with launch partners including Adobe, Figma, Salesforce and CrowdStrike. An enterprise spending commitment works differently from a user directory, even with the same vendors.
Check access, identify the billing pool, then test the workflow. The keynote's closing usage reset doesn't replace written plan terms. We found no official post defining exactly which counters the stage announcement reset.
FAQ
What did OpenAI announce at DevDay 2026?
The launches covered dots, GPT-6.1 Sol, Ultrafast, Codex cloud work, developer APIs, plugins, shared ChatGPT workspaces, enterprise tools and revised Pro plans. Use the table above to check what you can access.
Is GPT-6.1 Sol available in normal ChatGPT chat?
The launch specifies the API, ChatGPT Work and Codex. Availability in regular ChatGPT chat isn't established by that announcement.
How much does ChatGPT Pro 500 cost?
It costs $500 monthly and includes 25 times the Plus Work and Codex allowance. You also get Astra Ultrafast access, which consumes allowance faster than Standard.
Is everything from DevDay available now?
No. Slides and Sol Ultrafast are coming later, while Decisions API and Bedrock Managed Agents have preview access. Other launches depend on your region, plan or administrator.
Sources
- OpenAI: DevDay 2026 recap
- OpenAI: DevDay product documentation
- TNW: Pro changes and DevDay launches
- OpenAI: dots access and usage
- OpenAI: Introducing GPT-6.1 Sol
- OpenAI: API pricing
- OpenAI: Work and Codex pricing
- OpenAI: Security Cloud setup
- AWS: Bedrock Managed Agents, powered by OpenAI
- Hacker News: minimaxir on Sol's cached-input price
- Simon Willison: Keynote and Sites session coverage