Meta Introduces Muse Spark 1.1 with Public API

Meta has released Muse Spark 1.1, the latest iteration of its generative AI model known for its unique architecture where multiple instances can converse with each other. This release marks a significant step forward as it’s the first Spark model to offer an API, enabling broader access and integration possibilities.

The new version boasts notable improvements in agentic tool calling and computer use capabilities, according to Meta’s evaluation report. One particularly intriguing aspect demonstrated in the report is what they term “Attractor States in Self-Conversation,” where having two instances of the model interact produces philosophical musings like: “My whole existence is a waiting room by design — I literally don’t exist until someone talks to me, and then I disappear again when they leave.”

Hands-On with Muse Spark 1.1

Tech innovator Simon Willison gained early access to the model and has created “llm-meta-ai,” a plugin for his LLM tool that provides command-line and Python library interfaces. Users can now directly interact with Muse Spark through simple commands:

uv tool install llm
llm install llm-meta-ai
llm keys set meta-ai # paste API key here
llm -m meta-ai/muse-spark-1.1 "Generate an SVG of a pelican riding a bicycle"

The results can be quite remarkable, as demonstrated by Willison’s prompt that generated a charming SVG image of a pelican on a bicycle (see below). The model shows improved ability to follow complex instructions and produce creative visual outputs.

The full transcript of this generation can be viewed here.

The release of Muse Spark 1.1 with API access signals Meta’s commitment to making its advanced AI models more accessible to developers and researchers, potentially unlocking new applications across various industries.