OpenAI DevDay 2026: What the Announcements Mean for Marketing Analytics Teams

OpenAI's DevDay 2026 recap lists more than 20 announcements, and for marketing analytics teams four of them matter: dots (always-on agents), GPT-6.1 Sol at a fifth of GPT-6 Astra's standard input and output prices, the Agents API with computer use, and MCP Events. All four make an agent act on your data, so the data underneath it becomes the constraint. This is the considered read, about a week after the keynote.

Announcements below come from OpenAI's own pages. What they mean for a marketing team is our analysis and is labeled as such.

Key Takeaways

• OpenAI held DevDay on September 29, 2026 in San Francisco and called it its biggest yet, with more than 20 announcements across ChatGPT, Codex and its models.
• GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens, one-fifth of GPT-6 Astra's standard input and output prices, according to OpenAI's launch post.
• Dots are always-on agents with their own cloud computer that OpenAI says connect to over 4,000 apps. They work on tasks without being prompted each time.
• The Agents API now supports computer use, and ChatGPT plugins can start automations from events through the proposed MCP Events specification.
• For analytics teams the bottleneck moves from the model to the data: an agent that acts on its own needs governed, consistent campaign, CRM and spend data to act on.

What did OpenAI announce at DevDay 2026?

OpenAI's recap groups the announcements into five sections: new ways of working, building with Codex and the API, customizing ChatGPT with plugins, people and AI working together, and doing more with a ChatGPT subscription. OpenAI also said ChatGPT is now a shared surface for humans and agents, reaching what it calls its 1.2B weekly users.

ASCII.jp's pre-event report said CEO Sam Altman would give the keynote, livestreamed free from 10 AM US time. Most items shipped the same day. A few, such as collaborative slides and the Decisions API broad release, were dated only as "in the coming weeks" or "in the coming days." Here is the part that touches analytics work.

• Dots. Always-on agents, available on Pro and Business Premium in eligible markets, with a beta for Enterprise, Edu and Healthcare when a workspace admin turns it on.
• GPT-6.1 Sol. An upgrade to GPT-6 Sol, available in ChatGPT Work and Codex and through the API as gpt-6.1-sol.
• Agents API with computer use. OpenAI runs the infrastructure while developers build the agent.
• Decisions API. A limited preview that returns answers to a fixed set of questions you define.
• MCP Events. Plugins can start automations when something happens in a connected app.

What are dots, and can they do analytics work?

OpenAI describes dots as "remarkably capable, always-on agents." Powered by GPT-6 Astra, each has its own cloud computer, learns from feedback, and works toward goals around the clock. OpenAI says they connect to over 4,000 apps through its plugin ecosystem and can be reached in ChatGPT, Slack or Teams.

OpenAI's own examples are close to marketing work. One has a dot learn a team's audience, positioning and creative standards, then revise launch materials when scope changes and return drafts for review. Another has a dot read a new interview transcript, find clip moments and draft social posts for approval. A third has a dot watch customer feedback and bring back tested fixes.

Our read: a dot is a worker with its own login, so the useful question is what it can see. A dot that drafts a weekly performance summary is only as good as the numbers it pulls. If spend, CRM revenue and campaign names sit in different places and conventions, it will confidently summarize a mess. OpenAI says Enterprise, Edu and Healthcare access is off by default until an admin enables it. For an analytics lead, that switch is the moment to decide which datasets a dot may touch.

How does GPT-6.1 Sol pricing change the cost of analysis?

For a team that runs agents on a fixed monthly plan, the useful number is how much work the same budget buys. That is how our own marketing team now compares coding agents, and Sol's price is what made the comparison worth running.

OpenAI's GPT-6.1 Sol launch post lists three prices per million tokens. GPT-6 Astra is $10.00 input, $50.00 output and $1.00 cached input. GPT-6.1 Sol is $2.00, $10.00 and $0.10. GPT-6 Luna is $0.10, $0.50 and $0.01. OpenAI says Sol nearly matches Astra on agentic coding, computer use and professional work at one-fifth of Astra's standard input and output prices.

OpenAI reports one result that maps to analyst work. On GDP.pdf, a benchmark that measures how accurately models answer professional questions using complex PDFs with tables, charts, diagrams and fine print, it says GPT-6.1 Sol scores higher than Opus 5.5 with fallbacks at less than half the cost per task. That is OpenAI's own evaluation, not a test on marketing data.

Availability is narrower than the price suggests. OpenAI states GPT-6.1 Sol is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and is not yet available in Chat.

Our read: cached input at $0.10 per million tokens matters most for analysts. A reporting agent re-reads the same schema, metric definitions and naming rules on every run, which is exactly the repeated context that cached pricing discounts. If you already compared tiers in our GPT-5.6 tier guide for marketing work, the same method applies: price a real task, not a benchmark.

What do the Agents API and Decisions API mean for reporting pipelines?

OpenAI says the Agents API now supports computer use, so developers can build agents that interact with software to finish tasks. It also brings Codex's multi-agent capabilities, tool search, tool calling and context compaction into an application, with OpenAI running the underlying infrastructure. It is available through the API and in Codex and ChatGPT Work on Pro 500 and Enterprise.

The Decisions API is smaller but fits analytics well. OpenAI's DevDay recap describes it as focusing Luna's intelligence on a set of user-defined questions with finite, pre-defined answers. Developers send text or images and get back an answer to classify content, route a request, or choose an agent's next action. It launched in limited preview with a broad release planned in the coming days.

Our read: classification is a quiet, high-volume job in marketing data. Mapping a messy campaign name to a channel or deciding whether a lead source is paid or organic are questions with a short list of allowed answers, and a model that must pick from your list is easier to audit than free text. That still needs a stable channel list and a log of every decision.

Why do MCP Events matter for marketing data?

OpenAI is adding support for the proposed MCP Events specification, so plugins can start automations when something happens in a connected app. OpenAI's example is asking ChatGPT to watch for new tasks on a project board, then read the linked documents and draft a plan while you are away. It is available on all plans.

Note the word "proposed": treat any automation built on it as something you may need to revise.

Our read: until now most marketing agents answered when asked. Events let a budget threshold, a conversion spike or a broken UTM pattern start the agent. That raises the bar on the data it reads when it wakes up, because an alert that fires on a bad number wastes more time than no alert. If you want to see how a model connects to marketing sources today, our MCP page states that Improvado gives AI agents governed access to 187+ data sources through MCP servers, usable from Claude, ChatGPT, Cursor or any MCP client.

What should a marketing analytics team do this month?

Four checks, in order, based on what OpenAI announced and where it said it is available:

1. Write down the datasets an agent may read. Name them, name the owner, and decide read-only versus write. Do this before an admin enables dots.
2. Freeze your naming rules. Channels, campaigns and metric definitions need one version. Agents repeat whatever convention they find.
3. Price one real task. Take your weekly report, run it on Sol, and compare cost and accuracy with your current setup, including cached input.
4. Keep a person on budget changes and log what the agent did, so any number in a board deck traces back to source rows.

For the wider picture of how teams already use AI on their own campaign data, see our guide to AI marketing analytics.

None of these requires a specific vendor. If the shared problem is that campaign, CRM and spend data live in different places, talk to an Improvado expert about connecting them in one governed layer before an agent starts reading them.

What did OpenAI not say?

OpenAI gives no firm date for GPT-6.1 Sol Ultrafast (the recap says "coming soon," the launch post "in the coming days"), no marketing-analytics benchmark, no detail on how dots handle a specific ad platform or warehouse, and no separate price for dots beyond the plans they are offered on.

A team that reads "connects to over 4,000 apps" as "works with our stack" will find integrations differ in what they can read and write. Test one connection at a time with your own data and record what the agent could see. Our guides on AI marketing prompts and org design for AI adoption cover the workflow side.

Where this leaves marketing analytics: the announcements raise what an agent can do without being asked, and lower the price of the model doing it. Neither fixes inconsistent data. The teams that benefit first will be the ones whose campaign, CRM and spend data is already unified and governed. To see that layer working with the models you already use, book a walkthrough with an Improvado expert.

Frequently Asked Questions

When was OpenAI DevDay 2026?

OpenAI held DevDay 2026 on September 29, 2026 in San Francisco. OpenAI's recap, published the same day, says it was its biggest DevDay yet, with more than 20 announcements across ChatGPT, Codex, its models and new ways of working with AI.

What is an OpenAI dot?

A dot is an always-on agent that OpenAI says is powered by GPT-6 Astra, has its own cloud computer and browser, and learns from feedback. It connects to apps through plugins, and you can reach it in ChatGPT, Slack or Teams. Availability depends on plan and market.

How much does GPT-6.1 Sol cost?

OpenAI lists GPT-6.1 Sol at $2 per million input tokens, $10 per million output tokens and $0.10 per million cached input tokens, as of its September 29, 2026 launch post. It is available through the API as gpt-6.1-sol and in ChatGPT Work and Codex, not yet in Chat.

Does DevDay change how marketing teams should build reporting?

It changes the order of work. Cheaper models and event-triggered agents make automation easier, so the limit becomes data quality, naming consistency and approval rules. Decide which datasets an agent may read, keep a person on budget actions, and log every automated number before enabling any agent.