Meta has officially entered a new era of artificial intelligence with the debut of Muse Spark, the first model in a next-generation series designed to bring what CEO Mark Zuckerberg describes as “personal superintelligence” to everyday users. Unlike traditional AI assistants that focus narrowly on productivity or search, Meta’s latest breakthrough aims to help individuals tackle what matters most in their personal and professional lives anywhere in the world. The announcement immediately lifted Meta’s stock, fueled by broader strength in technology shares and mounting excitement over the company’s aggressive pivot toward AI-first ecosystems.
Developed by Meta Superintelligence Labs, Muse Spark is rolling out today within the standalone Meta AI app in the United States, with a broader expansion to Facebook, Instagram, Messenger, and WhatsApp expected in the coming weeks. In a strategic move to diversify revenue, Meta plans to eventually open access to Muse Spark’s underlying technology through an API, allowing third-party developers to build on top of the model. Early applications include native multimodal reasoning, tool-use support, visual chain-of-thought processing, and multi-agent orchestration. Select users can already request a private API preview, while the model is publicly available at meta.ai and through the Meta AI app.
Perhaps most striking is Meta’s introduction of “Contemplating mode,” a novel architecture where multiple agents reason simultaneously to solve complex problems. In benchmark tests, this approach scored 58% on Humanity’s Last Exam and 38% on Frontier Science Research—significant achievements for a model of its class. Behind the scenes, Meta completely rebuilt its pretraining stack over nine months, introducing sweeping improvements to model architecture, optimization, and data curation. The company claims Muse Spark achieves the same capabilities as its predecessor, Llama 4 Maverick, using more than an order of magnitude less compute, signaling a major leap in efficiency.
Beyond raw performance, Muse Spark is distinguished by its real-world applications, particularly in health and visual reasoning. Meta collaborated with over 1,000 physicians to curate specialized training data for health-related responses. The model can now generate interactive visual displays that explain nutritional content and even illustrate which muscles are activated during specific exercises. With Muse Spark, Meta is not just launching another large language model, it is laying the groundwork for an AI companion that reasons, sees, and helps in deeply personal ways.
