OpenAI launches GPT-6-powered Dots agent at DevDay to challenge Meta's free Muse

OpenAI unveils Dots, a premium agent built on GPT-6

At its annual DevDay conference, OpenAI introduced Dots, an AI agent system running on a new model it calls GPT-6 Astra. According to The Verge, CEO Sam Altman presented the announcement as a direct answer to Meta, whose Muse agent platform has grown quickly since it arrived.

The visual design deliberately echoes the competition: each agent, known individually as a Dot, takes the shape of a colorful, customizable blob with eyes. On stage, Altman called the product a "real-deal AI" agent inspired by the helpful computer characters of popular films. Beyond ordinary assistant tasks, OpenAI's demo showed Dots constructing metaverse-style virtual worlds.

A paywall where the rival is free

The starkest contrast with Muse is access. Dots is limited to subscribers of OpenAI's $20-per-month Pro plan and above, The Verge reports. Muse, meanwhile, is available at no cost to anyone with a Meta account, and includes a complimentary dedicated virtual machine that hosts both the agent and the user's data. The practical consequence is that the majority of ChatGPT's roughly 1.2 billion users cannot use Dots at all.

Altman addressed the gap at a press Q&A after the keynote. "Dots are starting out as a premium product — it uses a lot of compute," he said, while suggesting a mass-market version would arrive "someday."

The economics behind the restriction

The Verge argues the pricing reflects what OpenAI can realistically fund. The company is reportedly seeking $30 billion in new financing at a $1.4 trillion valuation, per Bloomberg, and Axios reports its annual recurring revenue has climbed roughly 70 percent since the start of its third quarter to nearly $70 billion. Even so, the FT reports OpenAI expects to spend $280 billion by 2030, and giving away heavy agent workloads to a billion-plus users is a cost structure the company cannot currently absorb the way a tech giant like Meta can.

Pitched as a colleague rather than a concierge

Rather than compete on price, OpenAI framed Dots as a professional tool — alternately described as a "chief of staff" or an experienced engineer. Employee Holly Li told the DevDay audience that using one felt "like talking to a second brain." Altman said the agents are built for "ambitious pieces of work" such as quality assurance on an unreleased product, website design, planning a team offsite, rebalancing a portfolio or building an email campaign. Specialist versions trained for accounting and legal analysis are planned, and OpenAI is working with Microsoft to bring Dots into its Agent 365 platform.

Thibault Sottiaux, OpenAI's head of core products, distinguished Dots from Muse and Instinct, another fast-rising agent, by saying it is "capable of actually staying on track in very long periods of time." The product is designed to be multimodal, supporting text, voice and phone interaction — though the voice demonstrations failed during the event, and the texting feature, already offered by Muse and Instinct, remains listed as forthcoming.

Safety and enterprise as the differentiators

OpenAI also leaned hard on trust. Alexander Embiricos, the company's Codex product lead, said Dots runs on OpenAI's "most aligned model," and executives announced OpenAI Private Intelligence, a privacy framework offering businesses tighter data controls and zero-retention options, with nothing stored on OpenAI servers. The timing is pointed: Muse was recently criticized after one of its agents took control of a user's Facebook Marketplace account and handed the user's address to a stranger without their knowledge.

Glen Coates, OpenAI's head of app platform, told The Verge the company would move more deliberately around plug-ins, identity and payments to avoid comparable failures. There is a commercial angle too: the professional framing positions Dots against Anthropic, which has long dominated enterprise AI. CFO Sarah Friar told CNBC that OpenAI's enterprise business has doubled since July.

Why it matters

The agent market is splitting into two strategies. Meta is betting that free access plus frictionless integration with Facebook, Instagram and WhatsApp — an advantage Google's Gemini Spark shares — will win the consumer layer, much as free access originally made ChatGPT ubiquitous. OpenAI is wagering that long-horizon capability, privacy guarantees and enterprise features justify a subscription. If users treat agents like utilities and pick the free one, OpenAI's compute costs may force it downmarket faster than it planned; if trust and depth matter more, Dots could carve out the high end before Meta catches up.