OpenAI unveils 'dots' to rival to Meta's Muse, plus a $500 monthly plan

OpenAI used its DevDay event in San Francisco to fire off more than 20 launches at once, but the biggest one was easy to spot: new desktop agents called “dots.” They’re meant to compete with Meta’s Muse and the Instinct assistant, and OpenAI is pitching them as always-available helpers that can be assigned work through text, Slack, and Microsoft Teams.

The pitch is simple enough. A dot can keep track of a project, prep for an appointment, or chase down a to-do list without waiting for constant hand-holding. It works across apps the user has allowed, and it only pings a person when something needs review. OpenAI says the idea is for the agent to do the grunt work first and interrupt only when it has to.

There’s a privacy catch, and it’s a big one. Dots remember preferences, past conversations, and decisions, and that kind of memory can mean a lot of data piling up over time. That may help explain why OpenAI is withholding them from places with stricter privacy rules, including the European Economic Area, Switzerland, and the UK. OpenAI is also trying to keep the agents on a leash: they have to ask before sensitive actions like changing a password, deleting data permanently, or installing software.

OpenAI is giving all Pro and Business Premium users one dot for free, with paid options coming later. It also plans to let teams of dots work together, and it’s building a shared area inside ChatGPT called ChatGPT Space where people can create documents, spreadsheets, and slide decks without exporting them elsewhere. OpenAI says multiple teammates and agents will be able to edit the same deck and leave comments.

The pricing push is just as aggressive. OpenAI introduced a $500-a-month plan with the highest usage allowance and faster Codex, while making the $200-a-month Pro tier less generous by cutting its included usage and message limit. That plan is also losing its five-hour usage cap, and OpenAI says model quality will keep improving. At the same time, it rolled out GPT-6.1 Sol for developers through the API, pricing it at the same level as GPT-6 Sol but claiming it gets close to GPT-6 Astra at a fifth of the price.