Meta Launches Muse Spark, Its First AI Model From Superintelligence Labs
Meta just dropped Muse Spark — the first AI model to come out of its newly formed Meta Superintelligence Labs, led by chief AI officer Alexandr Wang. This is the model Meta hopes will close the gap with OpenAI, Google, and Anthropic, and it represents a significant strategic shift for a company that previously championed open-source AI with its Llama series.
What Is Muse Spark?
Muse Spark is Meta's first reasoning model — meaning it can work through problems step-by-step, try different strategies when its initial approach fails, and even orchestrate multiple AI subagents to tackle complex questions simultaneously. Think of it as the brain upgrade Meta AI has been waiting for.
The model accepts voice, text, and image inputs but currently produces text-only output. Despite being described as "small and fast by design," it's capable enough to reason through complex questions in science, math, and health — areas where previous Meta AI models struggled to compete.
The Alexandr Wang Factor
This launch is inseparable from Meta's blockbuster move last year: spending $14.3 billion to acquire a 49% stake in Scale AI and hiring its co-founder and CEO, Alexandr Wang, as Meta's first-ever chief AI officer. Wang was tasked with building Meta Superintelligence Labs from scratch, and Muse Spark — originally codenamed "Avocado" — is the first tangible result of that bet.
The fact that Meta rebuilt its entire AI stack from the ground up in just nine months is remarkable. Wang's team didn't iterate on Llama — they started fresh with a new architecture designed specifically for consumer-facing AI.
What's Actually New in Meta AI
Muse Spark powers a redesigned Meta AI experience with several notable upgrades:
- Dual modes: Switch between quick answers and deep reasoning depending on the task
- Parallel subagents: Meta AI can now launch multiple agents simultaneously — for example, one drafts a travel itinerary, another compares destinations, and a third finds activities, all at once
- Multimodal perception: Snap a photo and Meta AI can identify products, read labels, and provide contextual information
- Health assistance: Developed with physicians, Muse Spark can help navigate health questions including image-based queries
- Visual coding: Create websites and mini-games from a text prompt
- Shopping mode: Discover products, get styling advice, and browse recommendations pulled from creators and communities
Where You'll Find It
Muse Spark is rolling out now in the Meta AI app and meta.ai in the US, with expansion to WhatsApp, Instagram, Facebook, Messenger, and Ray-Ban AI glasses in the coming weeks. That's a potential reach of over 3 billion users — a distribution advantage that no other AI company can match.
The enterprise AI market is consolidating fast, and Meta's decision to keep Muse Spark proprietary suggests the company views it as a competitive moat rather than an open ecosystem play.
The Open Source Question
Here's the interesting twist: unlike Meta's Llama models, Muse Spark is proprietary. Meta says it "hopes to open-source future versions of the model," but for now, it will be available in private preview via API to select partners only. This marks a notable departure from Mark Zuckerberg's previous open-source-everything stance and suggests Meta views Muse Spark as too strategically important to give away.
This strategic pivot mirrors broader trends in AI investment where companies are increasingly protecting their most capable models while still contributing to the open-source ecosystem with smaller releases.
What This Means for the AI Race
Meta's entry into the reasoning model arena intensifies an already fierce competition. With OpenAI preparing for its IPO, Google pushing Gemini hard, and Anthropic's Claude models gaining enterprise traction, Muse Spark gives Meta a credible seat at the table — backed by unmatched distribution across its family of apps.
The AI talent war continues to shape these competitive dynamics, and Meta's $14.3 billion bet on Alexandr Wang may prove to be one of the most consequential moves of 2025.
The next generation of Muse models is already in development, and if this "small and fast" version is any indication, Meta Superintelligence Labs is just getting started.
Frequently Asked Questions
What is Meta Muse Spark?
Muse Spark is Meta's first AI reasoning model, built by Meta Superintelligence Labs under chief AI officer Alexandr Wang. It powers the upgraded Meta AI assistant across WhatsApp, Instagram, Facebook, Messenger, and Ray-Ban AI glasses with capabilities including multimodal understanding, parallel subagents, and health assistance.
Is Muse Spark open source like Meta's Llama models?
No, Muse Spark is proprietary — a significant departure from Meta's previous open-source AI strategy with Llama. Meta says it hopes to open-source future versions, but the current model is only available in private preview via API to select partners.
How is Muse Spark different from other AI assistants?
Muse Spark's key differentiator is distribution — it reaches over 3 billion users across Meta's apps including WhatsApp, Instagram, and Facebook. It also features parallel subagent orchestration, strong multimodal perception for real-world images, and physician-reviewed health assistance capabilities.
Who is Alexandr Wang and why does he matter?
Alexandr Wang is the co-founder and former CEO of Scale AI, which Meta acquired a 49% stake in for $14.3 billion in 2025. He became Meta's first-ever chief AI officer and leads Meta Superintelligence Labs, the unit that built Muse Spark from the ground up in just nine months.