Muse Spark 1.1 is a multimodal reasoning model provided by Meta. It is designed not only for conversation, but also for agent-based tasks that involve planning and carrying out multiple steps based on a user’s objective. The model can process various types of input, including text, images, video, audio, and PDF files. It can also work with long documents and multiple sources while maintaining the context needed to continue a task. The model supports external tool use, function calling, web search with citations, structured output, computer interaction, and coding assistance. Given a goal, it can gather the necessary information, divide the work into smaller tasks, connect to external tools, and use intermediate results to determine the next step. Unlike open-weight models such as the Llama family, Muse Spark 1.1 is a commercial model accessed through Meta AI and the Meta Model API rather than downloaded and run locally. Its intended applications include document analysis, workflow automation, research, code generation and editing, computer-based tasks, and AI agents connected to external services.