Muse Spark 1.1 Handles Multimodal Computer Tasks

AIatMeta · x · 2026-07-09

Meta indicated that Muse Spark 1.1 retains fine details over long processes and handles visual and audio inputs in real execution environments, making it well-suited for vision-to-code generation, image/video captioning, and agentic computer use.

For example, the post noted that it can extract useful photos from a phone video, understand product details, and operate a browser to create a Facebook Marketplace listing for the user.

Related event: Meta Launches Muse Spark 1.1: A Low-Cost, High-Performance Agentic Model(133 posts)→

Original post →

More from Models

Models channel →