Meta Unveils Muse Spark, Its First AI Model From Superintelligence Labs

Meta has unveiled Muse Spark, the first major AI model from its newly formed Superintelligence Labs (MSL), marking a significant shift in the company's artificial intelligence strategy. The model, developed over nine months under the leadership of former Scale AI CEO Alexandr Wang, delivers frontier-level reasoning capabilities while requiring an order of magnitude less compute than its predecessor, Llama 4 Maverick.

Key Highlights

• Muse Spark is a natively multimodal reasoning model supporting text, image, and voice inputs
• The model achieves 58% on Humanity's Last Exam and 38% on FrontierScience Research in its Contemplating mode
• Artificial Analysis ranks it third overall, behind only Gemini 3.1 Pro and GPT-5.4
• Meta AI app surged to No. 5 on the US App Store with an 87% increase in daily downloads

A Ground-Up Rebuild

Muse Spark is not an incremental update. Meta rebuilt its entire AI stack from scratch, rethinking model architecture, optimization techniques, and data curation. The result is a model that matches the performance of its much larger predecessor while using dramatically less computational resources.

The model operates in multiple modes: an Instant mode for quick responses, a Thinking mode for enhanced reasoning, and an upcoming Contemplating mode that extends reasoning time for complex problems. In Contemplating mode, Muse Spark demonstrates what Meta calls "thought compression" — the ability to solve problems with progressively fewer tokens after initially extending its reasoning chain.

Multimodal and Agentic by Design

Unlike models that bolt on multimodal capabilities as an afterthought, Muse Spark was built from the ground up to process text, images, and voice natively. The model includes visual chain-of-thought reasoning, allowing it to analyze images and visual content as part of its problem-solving process.

The chat interface provides 16 integrated tools, including web search and browsing, Python code execution with scientific libraries, image generation in artistic and realistic modes, and the ability to spawn sub-agents for complex multi-step tasks. Meta also integrated semantic search across Instagram, Threads, and Facebook content, along with third-party account linking for Google and Outlook calendars.

Availability and Rollout

Muse Spark currently powers the Meta AI app and meta.ai website, with plans to expand across WhatsApp, Instagram, Facebook, Messenger, and Meta's Ray-Ban AI glasses. For now, the model is available exclusively in the United States.

The launch has already made a measurable impact: the Meta AI app jumped from No. 57 to No. 5 on the US App Store, with an estimated 46,000 downloads on launch day alone. The app has accumulated over 60.5 million worldwide installs.

A private API preview is available to select partners. Alexandr Wang has indicated plans to open-source future versions of the model, signaling a potential return to Meta's open-weight strategy.

Impact on the AI Landscape

Muse Spark positions Meta as a serious contender in the frontier AI race alongside OpenAI, Google, and Anthropic. Morgan Stanley analyst Brian Nowak called the launch "the first step in re-rating META," noting benchmark performance exceeded investor expectations.

The model's efficiency gains are particularly noteworthy. By achieving comparable capabilities with 10x less compute, Meta is demonstrating that raw scale is not the only path to frontier performance — a finding that could reshape the economics of AI development across the industry.

What's Next

Meta plans to roll out Muse Spark to its full ecosystem of apps in the coming weeks, potentially putting frontier AI capabilities in front of billions of users worldwide. The company has also signaled investments in health applications, having collaborated with over 1,000 physicians to curate training data for factual health responses.

With capital expenditures projected at 115 to 135 billion dollars for 2026 — nearly double last year — Meta is clearly betting that AI will define its next decade. Muse Spark is the opening move in that strategy.

Source: Meta Newsroom