In response to competitive pressures from companies like OpenAI and Anthropic, Meta has recently unveiled Muse Spark, the initial product of its newly established Meta Superintelligence Labs. This unveiling marks a significant investment of billions into AI research and development, aiming to create a superintelligent AI system. The model has been described as multimodal and beneficial for various Meta products, including the Meta AI app and services on platforms such as WhatsApp, Instagram, and Facebook. Muse Spark is distinctive as it is designed to reason step by step through complex questions in fields such as science, math, and health, contrasting with Meta's prior models which offered instant answers based on training data. Following extensive evaluation under its updated safety framework, the model showed promising results, albeit with performance that is competitive rather than superior to existing models from rivals. Currently, it has achieved a score of 89.5% on the GPQA Diamond benchmark, which indicates PhD-level reasoning skills. The development of Muse Spark comes in the wake of Meta's sweeping organizational changes since the Llama 4 incident, which resulted in an overhaul of its AI strategy. This included the hiring of Alexandr Wang, the CEO of Scale AI, who now serves as Meta’s chief AI officer. The collaboration with over 1,000 physicians for training data underscores its potential use in health-related applications, raising important considerations about safety and privacy. Anticipating the integration of Muse Spark into various Meta platforms, the company aims to gradually roll out features that incorporate richer, more visual content into responses. This model represents a critical first step on what Meta describes as a 'scaling ladder' for future AI models, illustrating the company's commitment to both expanding its AI capabilities and ensuring safety in their deployment.