Meta Launches Muse Spark: A New AI Model from Its Elite Superintelligence Labs
In a significant move to catch up with rivals, Meta has officially released Muse Spark, the inaugural artificial intelligence model from its high-profile Meta Superintelligence Labs (MSL). This development marks a pivotal shift in the company's AI strategy, moving away from its previous open-source approach with Llama models to a closed, proprietary system.
The Genesis of Meta Superintelligence Labs
Mark Zuckerberg established the Superintelligence Labs team last year under the leadership of Chief AI Officer Alexandr Wang. This decision came after a series of setbacks that left Zuckerberg frustrated with Meta's lag behind competitors like OpenAI, Anthropic, and Google. In June 2025, Zuckerberg hired Wang as part of a massive $14 billion investment into Scale AI, signaling a deep commitment to revitalizing Meta's AI capabilities.
Since then, Zuckerberg has poured billions into acquiring top AI talent and has committed tens of billions more to infrastructure, including advanced data centers, to bridge the gap with industry leaders.
Introducing Muse Spark: A Multimodal Reasoning Powerhouse
In a social media announcement, Alexandr Wang revealed, "Today we're releasing Muse Spark, the first model from MSL. Nine months ago, we rebuilt our AI stack from scratch—new infrastructure, new architecture, new data pipelines. Muse Spark is the result of that work, and now it powers Meta AI."
The company elaborated in a blog post, stating that Muse Spark is the initial model in the new Muse series, which adopts a deliberate and scientific approach to model scaling. Each generation is designed to validate and build upon the previous one before scaling up. This first model is intentionally small and fast, yet it possesses the capability to reason through complex questions in science, math, and health.
Shengjia Zhao, Chief Scientist at Meta MSL, expressed excitement, writing, "Excited to share what we’ve been building at Meta Superintelligence Labs! We just released Muse Spark, our first AI model. It's a natively multimodal reasoning model and the first step on our path to personal superintelligence. We've overhauled our entire stack to support scaling, and this is just the beginning."
Key Features and Capabilities of Muse Spark
Muse Spark is engineered to assist users with everyday tasks through advanced multimodal reasoning. For example, it can calculate the calories in a meal from a photo or superimpose an image of a mug on a shelf to visualize its appearance. The model also introduces Contemplating mode, which runs multiple agents simultaneously to enhance reasoning power. In practical terms, while planning a family vacation, one agent can draft a travel itinerary while another searches for kid-friendly activities.
The model powers the Meta AI chatbot and offers three reasoning levels: "Instant" mode, "Thinking" mode, and "Contemplating" mode, with the latter providing research-grade responses. While Muse Spark excels in science, health, and math queries, it is reportedly weaker at coding tasks.
Internally known as Avocado during its nine-month development, Muse Spark represents a refresh of Meta's AI strategy. Unlike the open-source Llama models, Muse Spark is a closed model, with its design and code kept private—a decision influenced by Wang's advocacy for closed systems. The model was trained using several third-party open-source models, including Qwen from Alibaba Group Holding Ltd., as well as models from OpenAI and Google.
Availability and Integration Across Meta Platforms
Muse Spark will initially be available on the Meta AI app and website. In the coming weeks, it will replace the existing Llama models powering chatbots on WhatsApp, Instagram, Facebook, and Meta's smart glasses. The Meta AI chatbot will remain free for users, though the company is considering subscription fees in the future.
The revamped Meta AI with Muse Spark will also feature a Shopping mode, designed to help users buy clothes or decorate rooms by drawing inspiration from creators and communities across Meta's apps.
Meta's Catch-Up Game in the AI Race
Despite using generative AI to boost its advertising business and improve efficiencies, Meta has yet to make a significant impact in the AI model market. Competitors like OpenAI and Anthropic are now collectively valued at over $1 trillion, and Google's Gemini technology is rapidly gaining traction.
A Meta executive acknowledged to Bloomberg that Muse Spark isn't as capable in some areas as ChatGPT, Claude, or Gemini, but emphasized that it's an early step in the company's execution. The model is described as "an early data point on our trajectory," with several larger models already in development.
Organizational Strategy and Future Outlook
Meta has aimed to keep its AI division agile by granting researchers autonomy and minimizing its typically management-heavy structure. Wang oversees around 100 direct reports, fostering a nimble environment for innovation.
The company is betting that applying AI to everyday personal tasks will boost engagement with its over 3.5 billion users across social media platforms, potentially giving it an edge over rivals with smaller reach. As Meta continues to invest heavily in AI, Muse Spark lays the foundation for future advancements, with the next generation already underway.
