OpenAI DevDay 2026: What Is Live, What Is Only Announced, and What to Do First

OpenAI DevDay 2026 took place on September 29, and most of what OpenAI showed is already live. A few of the items agencies will care about most are not: GPT-6.1 Sol Ultrafast, Collaborative Slides and Private Inference are still on the way.

The official OpenAI DevDay 2026 recap lists more than 20 updates across ChatGPT, Codex and the API. This post sorts the ones that matter to agencies and to the nonprofits, associations, colleges and public agencies they build for. For each, it covers which plans get it, whether it is live, and what to do first. Status is as of October 5, 2026.

The short version: GPT-6.1 Sol is the model to test now. It costs one-fifth of GPT-6 Astra per token and is live in ChatGPT Work, Codex and the API. The new Pro 500 plan is for heavy individual users, not most teams. Codex Security Cloud and code review are the most useful new tools for web teams. Bedrock Managed Agents, the Decisions API and Private Inference are previews or plans, so do not build client work on them yet.

The day before: GPT-6.1 Astra was pulled

OpenAI DevDay 2026 opened one day after OpenAI pulled its next top model. On September 28, OpenAI canceled the planned October release of GPT-6.1 Astra. The Register quotes OpenAI’s head of safety systems saying the model "didn’t quite meet the bar in terms of staying within scope and authorization."

GPT-6 Astra stays available. We covered the decision in OpenAI pauses GPT-6.1 Astra. For DevDay, it means Sol, the cheaper model line, carried the model news.

OpenAI DevDay 2026 at a glance

GPT-6.1 Sol and the Ultrafast tier

OpenAI’s GPT-6.1 Sol announcement says it is live for Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex. It is "not yet available in Chat." Developers call it as gpt-6.1-sol.

The API price is $2 per million input tokens and $10 output, the same as GPT-6 Sol. GPT-6 Astra costs $10 and $50. OpenAI says GPT-6.1 Sol gets close to Astra on coding and professional work. Those are OpenAI’s own benchmarks, so test on your own tasks.

Ultrafast is a premium speed tier. The recap says GPT-6 Astra Ultrafast is "available today in the API and in ChatGPT Work and Codex on Pro 500 and Enterprise plans." OpenAI quotes up to 8 times faster output in Codex and up to 6 times in the API.

GPT-6.1 Sol Ultrafast is not live as far as we can confirm. The recap says it is "coming soon." OpenAI’s API changelog through October 5 adds Ultrafast only for gpt-6-astra. The API pricing page lists Astra Ultrafast at $60 input and $300 output, six times standard.

What to do first: if you use GPT-6 Sol or Astra in client tools, run a week of real work through gpt-6.1-sol. Our GPT-6 Sol and Luna post covers the pricing logic behind this model line.

Pro 500 and the Pro 200 change

The recap describes Pro 500 as "25 times the ChatGPT Plus allowance" with access to Ultrafast. It does not state a price. Fortune reports it at $500 a month, and OpenAI’s Pro tiers help page lists the same figure. In ChatGPT, Ultrafast is available only on Pro 500.

The quieter OpenAI DevDay 2026 change affects Pro 200. OpenAI’s help page says Pro 200 is moving to a lower included allowance. If you had an active Pro 200 plan between September 22 and 10 a.m. Pacific on September 29, you keep the old allowance through October 29. The price stays at $200 a month.

What to do first: if staff hold Pro 200 seats, check their usage before October 29. Do not move anyone to Pro 500 for Ultrafast unless slow responses are blocking paid work.

Dots, ChatGPT Space and Pages

Dots are always-on agents that keep working on a goal between chats. They are live on Pro and Business Premium, with a beta for Enterprise, Edu and Healthcare that admins must turn on. We covered them in What Are OpenAI’s Dots?

ChatGPT Space is a shared area for a team to work with ChatGPT. Pages are documents built for people and agents to edit together. Both are live on Pro, Business and Enterprise, on desktop and web. On mobile you can find, read and share pages, but creating and editing are "coming soon." Collaborative Slides are due "in the coming weeks."

What to do first: if you run a Business or Enterprise workspace, check who can create shared spaces before staff start moving client documents into them. Our overview of Chat, Work and Codex explains how these areas fit together.

Codex changes at OpenAI DevDay 2026

Codex got four updates that matter to web teams. If the product names are confusing, start with What Is OpenAI Codex?

Cloud environments. OpenAI’s Codex cloud docs say you pick repositories, add an install script, and decide on internet access and allowed domains. Each task gets its own workspace and keeps working while your computer sleeps.

Code review. GitHub reviews are generally available, and GitLab support is in preview. Codex can review a pull request before you open it and post comments on GitHub.

Codex Security Cloud. You install it as a plugin, connect GitHub, and run a one-time scan or continuous scanning of new commits. Findings come with remediation guidance, and "Fix with Codex" can open a proposed patch.

What to do first: run a one-time Security Cloud scan on one client theme or plugin repo you maintain. Review the findings by hand before you trust the patches.

API additions at OpenAI DevDay 2026

Decisions API. OpenAI says it focuses GPT-6 Luna on "a specific set of user-defined questions with finite pre-defined answers." A typical use is routing a form submission to one of five queues. It is in limited preview, and the recap promised a broad release "in the coming days." The changelog had no broad release as of October 5.

Computer use in the Agents API. Agents can now work in an OpenAI-hosted browser. The computer use guide says the browser needs the user’s approval for each new website. The Agents API itself is in public beta.

Bedrock Managed Agents. This is OpenAI’s agent setup running inside AWS, with each agent getting its own AWS identity and permissions. Amazon’s product page labels it a preview.

What to do first: if a client already runs on AWS, join the Bedrock preview for testing only. If you still have code on the retired Assistants API, our Assistants API shutdown post covers the move to newer APIs.

The OpenAI Marketplace

The Marketplace lets eligible Enterprise customers apply part of an existing OpenAI commitment to approved partner software. The Marketplace page lists more than 30 partners, including Adobe, Salesforce, HubSpot and Figma.

Eligibility depends on your OpenAI agreement and the partner product. Payment goes to the partner directly, and OpenAI counts it against your commitment. For now, Enterprise customers can express interest.

What to do first: if a client has an OpenAI Enterprise commitment and also pays for HubSpot or Salesforce, ask OpenAI whether some of that spend can count toward it.

Private Intelligence: Zero Data Retention and Private Inference

Private Intelligence is OpenAI’s name for two data protection options announced at OpenAI DevDay 2026. The first is Zero Data Retention with Private Safety Processing. OpenAI says it allows automated safety reviews "without giving OpenAI personnel access to the underlying content."

OpenAI’s setup guide says organizations already approved for Zero Data Retention can turn it on in the API console. You connect your own storage bucket on AWS, Azure or Google Cloud, and content there expires after 30 days.

Private Inference is the second option. The recap calls it a preview "coming this fall" that combines confidential computing with verifiable controls. OpenAI has an interest form but no date.

What to do first: for clients with sensitive member or student data, confirm whether their API projects are approved for Zero Data Retention. Do not promise Private Inference in a proposal until it ships.

Frequently Asked Questions