Introducing Muse Spark: Meta’s New AI Model Designed to Prioritize People and Context
Meta AI has announced the launch of Muse Spark, its inaugural model in a new series of large language models developed by Meta Superintelligence Labs. This advanced AI strives to create a more personalized, intelligent assistant capable of supporting users across various platforms and tasks.
*A New Benchmark in AI Development**
Over the past nine months, Meta Superintelligence Labs has undertaken a comprehensive rebuild of its AI infrastructure, emphasizing speed and scientific validation. Muse Spark marks the beginning of this new series, which follows a deliberate approach to model scaling. Each iteration is designed to validate and enhance the previous one before progressing to larger models.
Muse Spark is engineered to be small and fast, yet capable of reasoning through complex questions in science, mathematics, and health. It serves as a robust foundation, with the next generation already under development. Currently powering the Meta AI app and the website meta.ai, Muse Spark is built to support complex reasoning and multimodal tasks, paving the way toward personal superintelligence—a tool that can assist anyone, anywhere, with the matters that matter most.
**Enhancements to Meta AI**
The Meta AI app and meta.ai are receiving an upgrade that introduces a refreshed interface and expanded capabilities. Users can now switch seamlessly between modes tailored for quick answers or deep reasoning. Meta AI can launch multiple subagents simultaneously to tackle complex queries—for example, planning a trip by drafting itineraries, comparing destinations, and finding kid-friendly activities all at once—delivering faster, more comprehensive answers.
**Seeing and Understanding the World**
One of Muse Spark’s standout features is its strong multimodal perception, enabling Meta AI to see and interpret what users are viewing. For instance, users can snap a photo of an airport snack shelf, and Meta AI can identify and rank snacks based on protein content—eliminating the need to squint at labels. It can also scan products and compare options visually.
This perception capability extends to health, where Meta AI can assist with health-related questions involving images and charts, providing more detailed responses. Developed in collaboration with physicians, this feature aims to enhance the AI’s usefulness in navigating health concerns.
**Creative and Practical Applications**
Muse Spark excels at visual coding, allowing users to generate websites and mini-games from simple prompts. Whether it’s creating a dashboard for a surprise party, developing a retro arcade game, or launching a flight simulator, Meta AI makes it accessible and shareable.
**Personalized and Contextual Assistance**
Meta AI is also enhancing its ability to help users discover fashion, home decor ideas, and shopping options by drawing from styling inspiration and community content across Meta’s platforms. When researching travel or trending topics, Meta AI now surfaces relevant community posts and local insights, enriching the user experience with real-world context.
**Looking Forward**
The upgraded Meta AI experience—featuring Instant and Thinking modes—is rolling out across the United States today and will expand globally in the coming weeks. The features will be integrated into Instagram, Facebook, Messenger, WhatsApp, and Meta’s AI glasses, where perception capabilities will be even more powerful.
Meta plans to offer access to this technology via private preview through API partnerships, with intentions to open-source future versions. As the models evolve, Meta commits to enhancing safety and privacy protections, ensuring responsible development.
Meta envisions a future where AI is deeply integrated into daily life, not just answering questions but truly understanding the user’s world—advancing toward the goal of personal superintelligence built on relationships and context.