Meta Platforms has pivoted from open-source evangelist to closed-model competitor with Muse Spark, its first major AI model since a $14 billion investment in Scale AI talent and infrastructure last year. The proprietary system marks a strategic reversal for the company that previously championed open-source AI through its Llama family.
Muse Spark will power Meta's digital assistant across Facebook, Instagram, WhatsApp, Messenger, and Ray-Ban Meta AI glasses starting in coming weeks.
The model was developed by Meta Superintelligence Labs, the new research team led by Chief AI Officer Alexandr Wang that Zuckerberg established after frustration with Meta's lagging position against OpenAI, Anthropic, and Google.
Meta shares jumped 6% following Wednesday's announcement as investors responded to the company's aggressive catch-up effort. The Facebook founder has spent billions on AI talent and committed tens of billions more on data center infrastructure this year alone.
The company plans to eventually offer paid API access to Muse Spark for third-party developers, creating a new revenue stream alongside its advertising business. Currently only select partners can access the private API preview.
Unlike Meta's previous open-source approach with Llama models, Muse Spark will remain proprietary though the company says there is "hope to open-source future versions." Wang personally favors closed models according to internal sources.
Meta acknowledges Muse Spark isn't as capable as ChatGPT or Gemini in some areas but calls it "an early data point on our trajectory" with several larger models in development.
The model was built over nine months using training techniques that include distillation from third-party models including Alibaba's Qwen as well as OpenAI and Google systems.
Users of the standalone Meta AI app will access different reasoning modes based on query complexity: Instant for quick answers, Thinking for intermediate tasks, and Contemplating for research-grade responses that use parallel AI agents. A Shopping mode will help users find clothing or furniture using inspiration from creators across Meta's apps.
The global generative AI market is projected to grow more than 40% annually from about $22 billion in 2025 to nearly $325 billion by 2033 according to Grand View Research. Meta's AI-related capital expenditures for 2026 will range between $115 billion and $135 billion, nearly double last year's spending.
