Meta is reentering the AI race with a new model called Muse Spark

Meta Superintelligence Labs has officially launched Muse Spark, its first major AI model since Mark Zuckerberg's multi-billion dollar overhaul of the company's AI division. The model is now live in the Meta AI app and website for US users, with a full rollout planned for WhatsApp, Instagram, Facebook, Messenger, and Meta's smart glasses in the coming weeks. As the inaugural model in the new Muse series, it's designed to be deeply integrated across Meta's product ecosystem, similar to Google's approach with Gemini. Meta is also offering API access to select partners in a private preview.

Technically, Muse Spark introduces several key features aimed at improving user experience and capability. It can run multiple AI sub-agents to handle queries more efficiently and supports multimodal input, processing both text and images—a critical function for Meta's AI-powered camera glasses. Users can toggle between a fast 'Instant' mode and a more deliberate 'Thinking' mode for complex reasoning. A significant focus is on handling complex questions in science, math, and health, with Meta highlighting its multimodal perception's ability to analyze health-related images and charts. This positions Muse Spark as a direct competitor to specialized health AI like OpenAI's ChatGPT Health and Anthropic's Claude for Healthcare.

The launch marks Meta's strategic re-entry into the high-stakes AI race following the disappointing release of Llama 4 in 2025. Muse Spark is billed as an 'early data point' for the Muse series, with Meta already developing larger models and expressing intentions to open-source future versions. The company also teased future features that could cite recommendations and content from across its social platforms, suggesting a deeply personalized AI experience. This move signals Meta's commitment to embedding advanced AI as a core utility across its entire family of apps and hardware.