Meta's Muse Outpaces ChatGPT on Early Mobile Metrics
Meta's Muse has outpaced ChatGPT on early mobile engagement in the United States. Appfigures estimates put Muse at 642,000 U.S. mobile daily active users compared with 231,000 for ChatGPT at the same point after launch, according to reporting published Sept. 21 TechCrunch.
The gap extends to installs. Muse saw 2.8 million total installs globally in its first 12 days. On a like-for-like iOS comparison limited to the U.S. and Canada, Muse recorded 1.8 million downloads in its first twelve days versus 1.3 million for ChatGPT over the same window.
That comparison matters because the launch footprints were not identical. Muse launched on Apple's App Store and Google Play in the U.S. and Canada only, while ChatGPT launched on mobile globally on iOS only. The Sept. 21 estimates normalize for that difference by isolating iOS in the U.S. and Canada. On iOS alone, Muse also led on engagement with 359,000 daily active users at the same point after launch, above ChatGPT's figure.
Chart position tracked the same trajectory. Muse moved from No. 2 overall on the U.S. App Store immediately after launch to No. 1, ranking higher than ChatGPT.
Installs, averages and distribution
Earlier snapshots had framed a closer race on velocity alone. Reporting published Sept. 20 cited ChatGPT averaging 87,000 downloads per day over its first eight days, against Muse averaging roughly 73,000 U.S. downloads per day over its first 10 days Yahoo. A Sept. 10 estimate had put ChatGPT at an average of 83,300 daily downloads around its launch, with Muse downloading more than 83,000 times on iOS in the United States shortly after launch.
The 12-day totals change the picture. Muse reached No. 1 with lower early daily averages in the U.S., which points to sustained install flow and broader device coverage rather than a single-day spike.
Distribution did the heavy lifting. Appfigures found over 95% of Muse's users are also Facebook users and 63% are Instagram users. For an expert audience, the implication is direct. Muse did not launch cold. It launched into an existing identity graph, notification surface, and app-switching habit.
What Muse is built to do
Meta describes Muse as its personal AI agent Meta. The company states Muse can browse the web, connect to users' apps, complete multi-step tasks, and keep working in the background with user direction and oversight.
The model underneath is Muse Spark. Meta describes Muse Spark 1.1 as a multimodal reasoning model built for agentic tasks Meta. The base Muse Spark design is natively multimodal with support for tool-use, visual chain of thought, and multi-agent orchestration.
Two product extensions are already attached to that stack. Muse Code is a coding agent powered by Muse Spark. Muse Image and Muse Video are the first media generation models developed by Meta Superintelligence Labs.
For practitioners, the vocabulary is familiar. Tool-use, background execution, and orchestration shift load from single-turn inference to persistent sessions with state, permissions, and retries across first- and third-party APIs.
The broader context here is that installs and daily active use measure different things, and agents widen the gap between them. A chatbot can succeed with short sessions and high return frequency. An agent that holds tasks in the background can generate fewer foreground opens while consuming more inference, tool calls, and storage. The 642,000 U.S. mobile daily active figure is therefore best read as a floor for demand, not a ceiling.
In my view, the number to watch next is retention after the Facebook-driven install wave. Cross-app ownership lowers acquisition cost. It does not guarantee repeated delegation of complex tasks. Agentic products live or die on permission grants, connector reliability, latency under multi-step planning, and graceful failure when a web flow or API changes. Early rank is cheap. Sustained delegation is expensive to earn.
Looking at what this means for builders, Muse normalizes a pattern enterprise teams already know from internal copilots. The client is thin. The work is asynchronous. Oversight is explicit. If Meta can keep task completion rates high without increasing user review burden, it will reset expectations for what a consumer assistant is supposed to do on mobile. That outcome would matter more than whether the first 12 days beat a 2023 launch curve. It would make background agents, with audit trails and app connectors, the default interface for routine mobile work.
