According to OpenAI, it has started rolling out "dots," a persistent AI agent product built on its GPT-6 Astra model, to Pro, Business Premium and Enterprise ChatGPT plans in "eligible markets" it did not name.
What a dot does
Per OpenAI, each dot runs on its own cloud computer, connects to more than 4,000 apps through OpenAI's plugin ecosystem, and is reachable through ChatGPT, Slack or Teams, including by voice. A dot is meant to keep working on a task across a stretch of time rather than waiting for a fresh prompt each step, and OpenAI says it learns a user's preferences and standards the more it is used. OpenAI's own examples are a developer's dot that turns customer feedback into tested pull requests, a product lead's dot that reworks launch materials as scope shifts and a scientist's dot that reruns analysis as new data arrives. All three end the same way: the dot brings back finished work and the person reviews or approves it.
The model behind it
OpenAI says dots run on GPT-6 Astra, the same model it named in September as the benchmark GPT-6.1 Sol was measured against on a coding task. That comparison made Astra a reference point; this announcement says Astra also powers dots. OpenAI has still not published benchmark results, technical specifications, or standalone pricing for Astra itself.
Specialist dots for company systems
Alongside the paid-plan rollout, OpenAI is previewing "specialist dots": versions with their own identity for access management, provisioned by a company's IT department, and built to plug into a company's systems of record for defined responsibilities. OpenAI gave no availability window or pricing for that preview.
What OpenAI did not say
The announcement does not name which countries count as "eligible markets," discloses no cost beyond the existing paid plans dots are rolling into, and gives no detail on how a dot decides when to act without checking in versus when it waits for a user's sign-off, beyond the pattern in its own examples of returning finished work for review or approval. OpenAI also has not described the security boundary between a dot's cloud computer and a user's own machine, beyond saying a user can "give your dot permission to connect and use your laptop."