Meta announced that its latest large language model, Muse Spark, developed by Meta Superintelligence Labs, is now rolling out to all of its core products — WhatsApp, Instagram, Facebook, Messenger, and the Ray-Ban Meta and Oakley Meta smart glasses. That means billions of users will get their hands on a new Meta AI that can see and orchestrate multiple agents at once.
The biggest underlying change this time is the introduction of parallel processing with sub-agents. Where AI used to handle questions one at a time, now when you throw a complex request at Meta AI, it can quietly spin up multiple agents to work in the background. Say you need to plan a family trip: one agent handles the itinerary, another compares destinations, a third searches for kid-friendly activities, and then they all sync up their results. The wait time for complicated tasks shrinks dramatically.
On the interaction front, Muse Spark finally gives Meta real-time multimodal perception. With smart glasses or a phone camera, you can look at the world, ask questions naturally, interrupt, or switch languages mid-conversation. The model reads environmental information directly — like scanning an airport shelf to find the snack with the most protein, or helping you interpret a medical chart. It can also generate visual code: just give it a one-sentence prompt, and it'll whip up a custom webpage or mini-game.
To truly leverage its super-app ecosystem, Meta is baking AI deep into social and shopping experiences. In a group chat, you can summon a sidebar that lets the AI tap into the conversation context for private answers. A shopping mode can compare prices across Facebook Marketplace and the broader web, then map them out. While other AI labs are mostly fighting it out with standalone apps, Meta is turning its massive hardware and software traffic funnel into ubiquitous infrastructure.