Meta’s new AI model can control your desktop, mobile and browser. Here’s how it works

Meta’s newest AI model isn’t just answering questions — it’s designed to operate computers directly. Muse Spark 1.1, which the company made publicly available to developers on Thursday, is trained to use computer interfaces on desktop, mobile, or browser, and can delegate execution to sub-agents running in parallel while handling long-running tasks with a 1M token context window.

According to Meta, the system can write and debug software code, navigate external applications, analyse text, images and video simultaneously, and execute multi-step logic workflows with very little human supervision. The company is pitching it as a framework for “personal superintelligence,” describing it as its most capable model yet for real-world software coding and “agentic” tasks — complex digital operations an AI can handle autonomously.

The launch represents a strategic shift for Meta, moving the company from purely free, open-source AI technology into the paid enterprise market, where it will compete directly against Anthropic and OpenAI. Developers in the US can now access Muse Spark 1.1 in public preview on the new Meta Model API, with $20 in free trial credits offered before a switch to pay-as-you-go pricing.

The announcement itself came through an unusual channel: CEO Mark Zuckerberg posted about the launch on X, ending a three-year silence on Elon Musk’s platform. “The Meta Model API allows developers to build using Muse Spark for the first time. Our focus is on delivering strong agentic and multimodal models at very low cost,” Zuckerberg wrote.