Meta Launches Muse Spark — Its First Closed AI Model From Superintelligence Labs
Summary
Meta has officially launched Muse Spark, the inaugural AI model from its Meta Superintelligence Labs (MSL) division. Unlike Meta’s open-source Llama family, Muse Spark is a closed, proprietary model — a significant strategic pivot for a company that built its AI reputation on open weights.
Muse Spark is natively multimodal, capable of processing text, images, and visual context simultaneously. It powers Meta’s AI assistant (available at meta.ai and in the Meta AI app) with multiple reasoning modes: “Instant” for quick responses, “Thinking” for deeper problem-solving, and an upcoming “Contemplating” mode that orchestrates multiple AI agents for complex tasks.
Meta claims Muse Spark is competitive with Anthropic’s Claude Opus 4.6, Google’s Gemini 3.1 Pro, and GPT 5.4 on key benchmarks, particularly in language understanding and visual reasoning. A global rollout across WhatsApp, Instagram, Facebook, Messenger, and Meta’s AI glasses is expected in the coming weeks, with private API access available to select partners.
Source
Meta Blog · Simon Willison · The Guardian
Commentary
The closed-model move is the most interesting part of this announcement. Meta spent years positioning Llama as the open-source champion against OpenAI and Google’s closed offerings. Now they’re essentially running both plays — open Llama for developer mindshare and ecosystem lock-in, closed Muse for their own product stack where they can capture the full value.
The multi-agent “Contemplating” mode is worth watching. If Meta can ship reliable multi-agent orchestration across 3+ billion monthly users on WhatsApp and Instagram, they’ll have by far the largest deployment of agentic AI in history. That’s a moat that no API-first AI company can match — distribution wins again.
