Meta Platforms Inc.’s latest bid at generative artificial intelligence (AI) is yielding immediate dividends. Following Wednesday’s release of Muse Spark, the company’s flagship AI app skyrocketed from No. 57 to No. 5 on the Apple App Store in 24 hours, according to Appfigures data.
The model marks the first major release under Alexandr Wang, the former Scale AI CEO recruited to lead Meta’s superintelligence division. Positioned as a “significant upgrade” over Llama 4, Muse Spark features advanced multimodal capabilities, allowing users to process voice, text, and images.
From solving complex scientific equations to visual coding for mini-games, the launch represents a bold stab at disrupting the dominance of OpenAI and Anthropic. The surge follows Meta’s massive capital injection into the sector, including a $14.3 billion investment in Scale AI.
Muse Spark marks the inaugural release from the Meta Superintelligence Lab (MSL), established by CEO Mark Zuckerberg in June. The lab’s creation followed internal pivots and the strategic recruitment of industry heavyweights, most notably Scale AI founder Wang.
Analysts said the launch represents a significant turning point for Meta, which has faced past criticism regarding the competitive performance of its previous AI iterations.
In a Facebook post announcing the release, Zuckerberg described Muse Spark as a “world-class assistant” with specialized capabilities in visual understanding, health, shopping, and gaming. The model’s architecture utilizes autonomous subagents capable of multitasking to fulfill complex user requests.
For example, Meta demonstrated the AI’s ability to plan comprehensive family trips by simultaneously drafting itineraries, comparing geographical locations, and sourcing local activities. The model also features advanced computer vision; a user can photograph airport snacks, and the AI will instantly analyze the nutritional data to identify the highest protein options.
Wang, writing on X, noted that Muse Spark includes a “contemplating mode” designed for extreme reasoning. He claimed the model is now competitive with top-tier rivals such as Google’s Gemini Deep Think and OpenAI’s GPT Pro.
According to Meta’s internal evaluation documents, Muse Spark outpaces models from Anthropic, Google, and xAI in several key benchmarks, though it trails in others. In a departure from Meta’s traditional “open science” approach, the company confirmed that Muse Spark will not be open source, though it expressed hope to open-source future iterations.
The aggressive rollout comes as Meta significantly ramps up its infrastructure investment. The company recently projected its 2026 capital expenditure to reach between $115 billion and $135 billion, a massive leap from the $72.22 billion spent in 2025.
Wall Street appears convinced by the spending trajectory. Meta’s 2025 revenue hit $198.8 billion—a 22% year-over-year increase—and analysts are currently forecasting 2026 revenue to climb to $247.7 billion. With Muse Spark now powering the company’s core AI services, Meta is betting that “personal superintelligence” will be the primary engine for its next era of growth.
