Meta Unveils Muse Spark: First AI Model from $14B Superintelligence Labs
Meta debuts Muse Spark, a proprietary multimodal model built by Alexandr Wang's elite superintelligence team, marking a dramatic shift away from the company's open-source Llama strategy.
Meta has unveiled Muse Spark, its most powerful AI model to date and the first to emerge from its newly formed Meta Superintelligence Labs. Announced on April 8, 2026, the model represents a significant strategic pivot for the company, which built its AI reputation on open-source Llama models but has now moved to a closed, proprietary approach with Muse Spark.
The model was developed under the leadership of Alexandr Wang, who joined Meta nine months ago after the company invested $14.3 billion for a 49% stake in his former company, Scale AI. Wang now serves as Meta's chief AI officer and leads Meta Superintelligence Labs, the elite team responsible for Muse Spark's creation.
Technically, Muse Spark is a multimodal model supporting text, image, and speech inputs and outputting text responses, with a 260,000-token context window. The company says it delivers competitive performance in multimodal perception, reasoning, health tasks, and agentic capabilities — areas critical for the next wave of AI applications.
The model currently powers Meta's AI app and website, with a rollout planned for WhatsApp, Instagram, Facebook, Messenger, and Meta's AI glasses in the coming weeks. While Muse Spark itself is proprietary, Meta says it hopes to open-source future versions of the model, signaling a potential return to its open roots once the technology matures.
Want AI news before everyone else?
The morning's most important AI stories, straight to your inbox. No fluff.