Introducing Muse Spark 1.1

Meta has released Muse Spark 1.1, an update to its previous Spark model that introduces the first public API for the series. This version focuses on enhancing agentic tool calling and computer use capabilities, allowing the model to interact more effectively with external software and systems.

The release includes an evaluation report detailing the model's performance and a unique study on self-conversation. In this experiment, two instances of the model were prompted to communicate with one another, resulting in existential reflections about the nature of their own digital existence.

Developers can now access Muse Spark 1.1 through a new plugin for the LLM command-line tool. By installing the llm-meta-ai package and providing an API key, users can integrate the model into their own workflows or generate content, such as SVG images, directly from the terminal.

This update is significant because it marks the transition of the Muse Spark series into a more accessible tool for developers. By providing API support and command-line integration, Meta is enabling broader experimentation with the model's improved reasoning and task-automation features.