Meta's 816B Muse model reaches desktops — with a hundreds-of-gigabytes catch

Meta released Muse Spark Open, an 816-billion-parameter multimodal model with open weights deployable on high-memory desktop-class workstations requiring roughly 256-300GB of memory. Despite sparse activation lowering compute per token, full large checkpoints or compressed versions need substantial storage, shifting inference cost to specialized hardware setups and offering hosted access as a practical alternative.

This Wire brief sits within Fusion42's coverage of AI Frontier Models and AI Infrastructure.