Muse Spark is Meta’s inaugural product in the Muse family from Meta Superintelligence Labs. It is a natively multimodal reasoning model designed to leverage tools, perform visual and textual reasoning, and coordinate multiple agents. Targeting developers, researchers, and product teams, Muse Spark aims to support personal-scale AI workflows—from environment analysis to wellness-oriented assistance—through a scalable AI stack that Meta positions as foundational for future personal superintelligence. The launch also signals an overarching effort to scale AI capabilities across research, model training, and infrastructure, including investments in data-center capacity. Muse Spark is available today at meta.ai and within the Meta AI app, with a private API preview opening for selected users.
Muse Spark’s stated goal is to advance toward a personal superintelligence that can understand and interact with a user’s world, balancing perception, reasoning, and action across modalities and tasks. In practice, this means handling multimodal inputs (images and text), reasoning through problems, and orchestrating actions via tools and multiple agents. Meta highlights Contemplating mode as a major capability, enabling parallel agent reasoning to push performance on challenging tasks and making Muse Spark competitive with frontier models in certain reasoning scenarios. The rollout of Contemplating mode will be gradual, signaling a staged approach to expanding capabilities and use cases.
For readers and potential users, Muse Spark positions itself as an early-layer building block in a broader scaling effort, with initial focus areas including multimodal perception, reasoning, health-related tasks, and agentic workflows. Applications span both personal-use scenarios—where the model analyzes surroundings or supports wellness—and more technical workloads that involve coding workflows or complex decision-making that benefits from multi-agent orchestration.
