The race to build AI that does things for you, not just talks to you, now has two very different front-runners. Meta and OpenAI have each released an AI agent, and they are aiming at almost opposite ends of the market.
Meta’s Muse is free and built for the masses. OpenAI’s Dots is paid and built for the office. According to Bloomberg, the two approaches are likely to become similar over time.
Two launches, three weeks apart
Meta moved first. It launched Muse on September 8, 2026, offering a free tier backed by dedicated virtual machines.
That last detail matters more than it sounds. A virtual machine is essentially a private computer in the cloud. Giving each user one means the agent has its own workspace to click, browse and complete tasks without touching your actual device.
The early numbers suggest the strategy landed. Muse reportedly passed 6.6 million downloads shortly after its debut, with a 73% day-1 retention rate.
Meta is pushing Muse across its own platforms, including Facebook and Instagram.
OpenAI took a different path three weeks later. It launched Dots on September 29, 2026, pitching it squarely at enterprise users.
Dots runs on OpenAI’s GPT-6 Astra model. Access comes through paid plans that start at $100 per month.
OpenAI requires explicit user consent before Dots handles sensitive tasks.
OpenAI also has a sizable launchpad. The company reported over 35 million weekly users for its Codex and ChatGPT features by late September 2026, giving Dots a ready-made audience of people already using OpenAI tools for professional workflows.
Why the strategies look so different
Safety is the obvious challenge. An agent that can act on your behalf can also make mistakes on your behalf, and the consequences scale with how much access you give it.
Interoperability is the quieter issue. If every company’s agent speaks its own language, agents cannot easily work with outside services or with each other.
Meta is trying to address that gap through the Personal Agent Protocol, announced on October 6, 2026. Meta is participating in the initiative alongside other large industry players, with the goal of tackling interoperability challenges.
What this means for the agent race
The immediate takeaway is market segmentation. Consumers who want to try an AI agent without spending money have an easy entry point in Muse. Businesses that need guardrails and accountability have a clearer fit in Dots.
Early market reaction has reportedly favored Meta, with the company seeing stock gains as interest in its agent push has grown.
The bigger question is how long the split lasts. Bloomberg’s framing suggests the two products are likely to converge.
Meta’s free virtual machines are costly to run, and a 73% day-1 retention rate says nothing yet about week-four or month-six engagement. OpenAI faces a separate challenge: a $100 starting price limits casual experimentation, and its 35 million weekly users represent a base still to be converted to a premium agent product.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.