Meta Launches Muse Spark, Its First Proprietary AI Model in Break from Open-Source Roots

Meta has unveiled Muse Spark, the first AI model produced by its Superintelligence Labs division — and in a move that has drawn sharp reactions across the AI community, the model is entirely proprietary. The release marks a striking departure from Meta's longstanding commitment to open-source AI development, which defined its influential Llama model series.

Built by Alexandr Wang's Lab

Muse Spark is the debut product of Meta Superintelligence Labs, the research unit led by Alexandr Wang, the former Scale AI CEO who joined Meta roughly nine months ago as part of a reported $14 billion acquisition and hiring deal. The lab was tasked with pushing the frontier of AI capability, and Muse Spark represents its first public output. According to Meta, the model family includes smaller variants that are as capable as older midsize Llama 4 models while requiring "an order of magnitude less compute" to run — a significant efficiency gain that could reduce inference costs for enterprise customers.

The Open-Source Debate

The decision to keep Muse Spark proprietary has ignited controversy. Meta built enormous goodwill in the AI research community by releasing the Llama model series under permissive licenses, positioning itself as the counterweight to OpenAI and Google's closed approaches. Critics argue that the proprietary turn undermines Meta's credibility as a champion of "open science" and signals that competitive pressures are overriding philosophical commitments. Meta executives pushed back, saying the company hopes to open-source future versions of Muse models once safety evaluations are complete, though no timeline was offered.

Massive Capital Behind the Push

The launch comes amid an extraordinary investment cycle. Meta has committed between $115 billion and $135 billion in AI capital expenditure for 2026, spanning data center construction, custom chip development, and research operations. Wang's lab is understood to command a substantial share of that budget, reflecting CEO Mark Zuckerberg's conviction that achieving superintelligence-class models is essential to Meta's long-term competitive position. Whether Muse Spark delivers on that ambition — and whether Meta can maintain its open-source community while simultaneously building proprietary products — will be among the most closely watched storylines in AI this year.