OpenAI’s September 29 DevDay recap counts more than 20 major announcements. That number spans a model available through the API, agents restricted by plan and market, team features requiring administrator setup, previews and products promised for later. For developers and team leads, the useful question is which workflow can begin now.
The DevDay event brought personal agents, coding tools, API services and enterprise collaboration into one launch slate. The matrix below reconciles OpenAI’s recap with its product documents as of September 30. This is documentation-based reporting. BIG CHANGE has not tested these products hands on.
Personal agents, models and subscriptions
Announcement | Access and stage | What readers need to know |
|---|---|---|
Dots | Pro and Business Premium in eligible markets; admin-enabled beta for Enterprise, Edu and Healthcare, off by default. | The ongoing personal agent needs delegated access and review. Read OpenAI’s Dot announcement and our Dot permissions and pause guide. A product demonstration does not establish ordinary task reliability. |
GPT-6.1 Sol | API model | Standard API prices are $2 per million input tokens, $0.10 cached input tokens and $10 output tokens. OpenAI’s model article reports company evaluations, not observed reader outcomes. The Business and Enterprise/Edu notes clarify rollout. |
Ultrafast | GPT-6 Astra Ultrafast is available in the API and in Work and Codex on Pro 500 and Enterprise. GPT-6.1 Sol Ultrafast is coming soon. | OpenAI claims up to 8× faster token generation in Codex and 6× in the API for its premium speed tier. Those are vendor figures, not our latency measurements. |
Sign in with ChatGPT | Identity sign-in is global; shared plan usage is for Plus and Pro in participating tools. | Eligible partner usage draws on plan allowances and user controls. Availability depends on the partner. |
Pro 500 | Available now, according to OpenAI. | The tier offers a claimed 25 times the Plus usage allowance and includes Ultrafast. OpenAI describes the 25× figure as the Pro allowance relative to Plus; see the DevDay recap for the company’s stated comparison. |
“Always on” describes Dots’ mode of delegated work. It does not show that every assigned job will finish. Region, connected accounts, permissions and administrator enablement shape what an eligible agent can actually do. OpenAI’s recap and Dot announcement describe its claims and controls, without an independent field study.
Codex and API development
Announcement | Access and stage | Practical boundary |
|---|---|---|
Codex Cloud | Plus, Pro, Business, Healthcare, Education and Enterprise, subject to workspace cloud-access settings. | Tasks use isolated workspaces from published, reusable environments. Connect repositories, review prepared tools and access, then publish. The Cloud environments guide says computer and browser use are not currently supported in cloud environments. Our environment setup guide covers the steps. |
Refreshed Codex CLI; Code Review | Both listed for all plans. | The CLI adds voice steering and an |
Codex Security Cloud | Pro, Business, Enterprise and Edu on desktop and web. | OpenAI says it can scan GitHub repositories on demand or on a schedule, investigate findings and prepare fixes. Teams still decide what to merge. |
Decisions API | Limited preview on September 29, with broader availability planned “in the coming days.” | Developers define finite answer choices and provide text or image context for classification or routing. The recap describes the feature. The checked API changelog supplied no current Decisions API reference or confirmation of broad release; endpoint, schema and price remain undocumented in the sources checked. |
Agents API with computer use | Listed through the API and in Codex and Work on Pro 500 and Enterprise. | The computer-use guide documents an OpenAI-hosted browser and approval events handled by the application. This requires integration and oversight work. |
Bedrock Managed Agents | AWS says it combines OpenAI models, the Codex harness and Bedrock AgentCore, with runtime and inference inside AWS. AWS identity and permission controls determine reachable resources. |
OpenAI also introduced Private Intelligence. Its recap points to Zero Data Retention with Private Safety Processing, while Private Inference is a preview coming this fall. Those are separate stages. Buyers need the data-handling terms and deployment scope of the option they seek.
Plugins, shared work and enterprise purchasing
Announcement family | Access and stage | Setup and limit |
|---|---|---|
Plugin extensions; Plugin Creator, submission and discovery | Recap lists all plans, while extension docs say web extensions for Free and Go are coming soon and composer mentions are desktop only. | Extensions can add sidebar homes, interactive panels and file viewers. Users choose plugins and approve access. |
Sites plugins; MCP event automations | Sites plugins: Business, Enterprise, Healthcare and Edu, where Sites is available. MCP events: all plans, according to OpenAI. | The MCP Events guide requires MCP 2.0, persistent subscriptions and HTTPS callbacks; ChatGPT supports webhooks, not every draft-spec method. OpenAI’s plan guide says Sites is unavailable in the EEA, Switzerland and UK. Business notes say Sites app use needs organization approval and an admin setting off by default. Visitor access cannot be reused by background schedules. |
ChatGPT Space and Pages; collaborative slides | Space and Pages: Pro, Business and Enterprise. Mobile Space creation/editing is coming soon. Collaborative slides are promised for those plans in the coming weeks. | Shared context and documents are available where enabled; slides remain a future launch. |
Teams and shared tasks | Business and Enterprise, subject to workspace permissions and gradual rollout. | Admins configure approved connections; team owners enable them; the provider account defines accessible data and actions. The team-task documentation warns that announced features may not yet appear in every account. Our shared Team Task guide follows the setup. |
@ChatGPT in Slack and Microsoft Teams | Business and Enterprise after organization installation and enablement. | It uses administrator-connected tools and applicable permissions. Replies in shared conversations are visible to participants, according to the Enterprise/Edu notes. |
Meetings plugin | Beta in the macOS desktop app for Pro and Business; Enterprise is coming soon. | OpenAI says it saves summaries and action items to Space and deletes audio once notes are ready. |
Shareable profiles | Free, Go, Plus, Pro, Business and Enterprise. | Profiles collect Sites and plugins for others to find; skill sharing remains within a workspace. |
OpenAI Marketplace | Eligible Enterprise customers can express interest. Enterprise/Edu notes call it a managed beta. | Qualifying partner software can use part of an existing OpenAI commitment, subject to per-customer and product eligibility. Partners contract and invoice directly; no self-service checkout is offered. |
The team announcements shift recurring work from one person’s connected tools toward shared instructions and approved connections. The permission chain matters: an administrator allows the feature, a team owner chooses connections, and the provider account limits data and actions. The team-task guide says unattended runs cannot complete a new app sign-in and do not remove approval requirements for actions in connected apps.
The big change
DevDay grouped personal agents, code agents, API-hosted browser work and team automation under one launch count. Adoption depends on the particular surface, plan, market, administrator setting and connected-account scope, then on who reviews the result. OpenAI’s 20-plus figure describes the breadth of the announcement. It does not establish that 20 products are generally available or that these workflows have delivered measured gains in ordinary use.
Sources & further reading
- OpenAI DevDay 2026 recap: September 29 announcement list and OpenAI’s stated access stages. Product capability claims are the company’s.
- OpenAI DevDay event page: event date and setting, not a product availability reference.
- GPT-6.1 Sol and Dots: model pricing and evaluation context, and personal-agent rollout claims; neither establishes our own product test.
- Business release notes and Enterprise/Edu release notes: September 29 rollout, admin controls, connected-app scope and Marketplace beta.
- Codex Cloud environments, team tasks, Agents API computer use, plugin extensions and MCP Events: setup requirements and limitations.
- Using Codex with your ChatGPT plan: workspace cloud-access settings and regional Sites availability.
- AWS Bedrock Managed Agents: AWS’s Preview label. OpenAI API changelog: checked for a Decisions API reference, with none found at this review.
