Daily AI CatchupMeta Releases Muse Spark Open, 816B-Parameter Model
Meta's research arm released Muse Spark Open, an 816-billion-parameter multimodal open-weight model with 42B active parameters and a 1-million-token context window. It targets coding and long-horizon agentic tasks and, notably, can run locally across two Apple M5 Ultra Mac Studios using the MLX framework, meaning no cloud GPU cluster is strictly required to use it.
Why it matters
💻 Developer · A frontier-scale open-weight model you could plausibly run on prosumer Apple hardware via MLX, worth testing against your coding agent stack.
📦 Product · Open-weight models at this scale change what you can self-host instead of paying per-token API fees for agentic workloads.
🎨 Design · Long-context, multimodal open models like this will feed into more design tools that reason over large files and image references.
📈 Business · Running a near-frontier model on consumer-grade hardware lowers the cost floor for companies wanting data control without cloud lock-in.
🤔 Just Curious · A huge AI model that can run on two fancy Mac computers instead of a data center full of servers.
Sources: Muse Spark Open