Meta's Muse Just Beat ChatGPT at Its Own Launch Game

The Garden Wall Just Got Taller

Meta’s new AI app Muse has, according to estimates from Apptopia, attracted more mobile downloads in its first 12 days than ChatGPT did during its entire launch window. Daily active users are also higher. The numbers come from the U.S. and Canada — Meta’s home turf — but the implication stretches far beyond those borders.

Open models have been the darling of the AI industry since OpenAI launched ChatGPT in late 2022. The thesis was simple: accessibility wins. Any developer could build on top of Llama, Mistral, or whichever open-weights model happened to be the flavor of the month. Innovation would accelerate. Consumers would benefit.

Muse is a reminder that access alone doesn’t guarantee adoption.

The Numbers, Stripped Down

Apptopia’s analysis focused on iOS in the U.S. and Canada to make the comparison fair. ChatGPT launched only on iOS initially; Muse landed on both iOS and Android but only in those two markets. Stripping away platform and geography differences, Muse pulled 1.8 million downloads against ChatGPT’s 1.3 million. Daily active users in the same cohort: 642,000 for Muse versus 231,000 for ChatGPT.

That’s a 2.8-fold gap in engagement during a critical window — the time when most apps either find their footing or fade. For context, that DAU gap is comparable to what separates a breakout social app from a forgettable utility. It signals that users aren’t just downloading Muse; they’re opening it and staying.

Muse climbed from the No. 2 spot on the U.S. App Store immediately after launch to No. 1, surpassing ChatGPT itself, according to Business Insider. Appfigures had previously estimated the app crossed 1 million downloads. Meta hasn’t released its own figures, which means we’re working with third-party estimates — but all major tracking firms are pointing in the same direction.

The Real Weapon Was Never the Model

What’s striking here isn’t that Meta matched or beat ChatGPT on AI capability. It’s that Meta deployed its most underappreciated asset: its social graph.

Over 95% of Muse’s users are also Facebook users. 63% are Instagram users. The app integrates directly into WhatsApp. Meta didn’t need to convince strangers to try a new product; it needed to remind the people it already reached into hundreds of millions of homes to log in again.

This is the same playbook that built Threads. That app hit 500 million users because Meta gave it a frictionless on-ramp: one tap, existing credentials, no cold start. Threads didn’t win because it was better than Twitter at everything. It won because it removed every barrier between a Facebook user and a new app.

Muse is following the same script — but with a crucial difference. Threads was a social network competing against an existing behavior (posting). Muse is an AI assistant, a category where ChatGPT had already established deep habit formation. Beating ChatGPT at its own game requires more than convenience. It requires a different kind of switch cost to overcome.

The WhatsApp Factor and Second-Order Effects

The WhatsApp integration is where this gets strategically interesting. With over two billion users globally — and particularly strong penetration in India, Brazil, and parts of Europe — WhatsApp gives Muse a distribution channel that ChatGPT simply doesn’t have. This isn’t just about the U.S. and Canada numbers; it’s about where AI adoption is accelerating next.

For developers building on open models, the second-order effect is already visible. When a walled garden can ship an AI experience with a single tap, the value proposition of self-hosted or API-based open models shifts. Developers will still build on open models — the economics of that ecosystem are too entrenched now — but their addressable market narrows. The consumers who matter most, the ones who want AI baked into products they already use daily, are increasingly accessible only through proprietary gates.

OpenAI faces a strategic recalibration. Its advantage has always been capability plus openness. But if Meta’s distribution engine can neutralize the openness premium — if users are willing to trade model flexibility for app convenience — then OpenAI’s moat erodes from both sides. Google has been playing this game longer with Gemini embedded in Search and Workspace. Apple is quietly building AI into the iOS stack. Meta is now entering the ring with the heaviest distribution leverage of any competitor.

Pricing pressure is another likely consequence. Meta has signaled that Muse will be free, funded by its existing ad infrastructure. That makes it harder for OpenAI to justify a $20-per-month ChatGPT Plus subscription for features that are increasingly replicable inside walled gardens. The subscription model isn’t dead, but it needs to deliver demonstrably superior capability to survive against free, convenient alternatives.

Who Wins, Who Loses

Winners: Meta, whose AI division finally has a mobile app that matches the distribution reach of its social products. The integration with WhatsApp is particularly potent — it puts an AI assistant inside the messaging app that over two billion people use daily, across emerging markets where Meta’s other products have less traction.

Losers: The open-source AI narrative. Every time a walled garden ships an AI experience that outperforms the open alternative on adoption metrics, the argument for distributed model development loses a step. Developers will still build on open models. But consumers will keep choosing convenience.

Uncertain: What happens when the novelty fades. The 12-day window measures launch excitement, not retention. ChatGPT has had years to build habit. Muse has had twelve days. The real test — whether people return to Muse after the first conversation — hasn’t happened yet.

What Comes Next

If Muse sustains even a fraction of its launch momentum, Meta becomes the first company to turn AI assistance into a social platform advantage. That changes the competitive calculus for every AI startup that hasn’t secured a distribution deal. The question is no longer just “how good is your model?” but “who controls the gateway to your users?”

Open-weight models will continue to advance. Llama 3.3, Grok, and others will close the capability gap. But closing the gap between what an AI can do and who actually uses it is a different problem entirely — and Meta seems to have solved it first.

The open AI movement believed the best model would win. Muse suggests the best-distributed model might win instead. And in a market where the difference between 1.3 million and 1.8 million users in two weeks is the difference between a trend and a transformation, distribution may be the only metric that ultimately matters.