avoid.worksNotes, news & experiments in AI.
← Back to news

OpenAI DevDay 2026: what changed beyond chat?

OpenAI announced a new model, an ongoing AI assistant, and more places to work with AI. Here is what each one means.

Hover, click, or tap a dotted-underlined term for a quick explanation and a link to read more.

Imagine you are organizing a small event. You might ask a chatbot for an invitation draft, then copy it, check the guest list, and send the messages yourself. At DevDay on September 29, OpenAI presented tools aimed at more of that work: an assistant that can keep working on a task, a shared place for the project, and ways to connect the apps involved. The announcements cover different parts of a system, not one magic model that does everything.

A person asks for an event update. A dot checks permitted apps and prepares work; the person reviews an important action before it happens.
A possible workflow, not an automatic promise: the assistant needs access to the right apps, and you still check important actions.

What was announced?

Dots are OpenAI's new ongoing AI assistants. OpenAI says a dot can use its own cloud computer and apps you connect, work on several tasks, and bring results or decisions back to you. In the event example, it might look for a venue reply, draft an update, and ask before sending it. Dots run on GPT-6 Astra; they are a product built around a model, rather than a new model named “Dots.” You decide which apps they can access. OpenAI says background research uses read-only tools, while actions affecting accounts go through rules and review. Dots can still make mistakes, so review consequential work.

GPT-6.1 Sol is a model update. A model is the part of an AI system that processes a request and produces a response or action. OpenAI says this Sol version improves on GPT-6 Sol in coding, computer use, and professional tasks while approaching its more expensive Astra model on some tests. “One-fifth the price” compares standard API token prices with Astra, not the price of a ChatGPT subscription or every finished task. OpenAI's benchmarks use particular tools and test settings; try a task you can check before treating a score as a promise.

Codex and ChatGPT Space are work environments. Codex gained more ways to run coding tasks from a computer, phone, or cloud workspace. Space is a shared place for a team to keep pages and project knowledge with ChatGPT and a dot. Neither name refers to a new underlying model. They change where work happens and what context the assistant can use.

Plugins and developer tools connect these pieces to other software. OpenAI announced expanded plugin experiences, computer use in its Agents API, and a Decisions API for choosing from a fixed set of answers. Think of these as building blocks for apps, rather than buttons every ChatGPT user will see. The Decisions API is in limited preview, while collaborative slides are planned for the coming weeks.

What can you try now?

Availability depends on the product and plan. OpenAI says GPT-6.1 Sol is available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, and through the API; it is not yet in standard Chat. Dots are rolling out in eligible markets for Pro and Business Premium users. Enterprise, Edu, and Healthcare workspaces can try the beta when an administrator enables it. ChatGPT Space is available to Pro, Business, and Enterprise users on desktop and web, with some mobile features available. The company also announced a Pro 500 plan, but it is not required for all these features.

Why does it matter?

The useful question is no longer only “Which model is smartest?” It is also “What can this setup see, what can it do, and how will I check the result?” Start with a low-stakes task, such as drafting a project update from a folder you choose. Inspect the sources and the draft before giving an assistant permission to send messages, edit files, or spend money. A capable agent with access to the wrong information can still produce the wrong outcome.

Sources & further reading