Meta unveiled a new artificial-intelligence model dubbed Muse Spark, the first product from its Superintelligence Labs led by former Scale AI chief Alexandr Wang, as the company races to close the gap with OpenAI, Anthropic and Google. Framed as a smaller, faster system with “competitive performance,” Muse Spark will power Meta’s assistant across its apps and devices and be offered to developers through a paid API after a limited preview. The launch marks a shift from Meta’s recent open-source posture: Muse Spark is proprietary, though the company says future versions could be opened. Meta says revamped training techniques and infrastructure allow Spark to match prior midsize Llama 4 capabilities with far less compute, and it is targeting multimodal reasoning, health, and agentic tasks with new modes for complex queries and shopping. The push comes as Meta plans $115 billion to $135 billion in AI-related capital spending in 2026 and seeks new revenue streams from its model portfolio. Shares rose 6.5% Wednesday alongside a broader market rally.
Related articles:
– Introducing Meta Llama 3
– Generative AI Market Size, Share & Trends Report
– Amazon Bedrock
