The Meta Muse Spark OpenAI model brings advanced coding and autonomous agent capabilities to Apple’s Mac Studio computers. Meta’s new open-weight system contains 816 billion parameters and supports a 1-million-token context window. However, running it locally requires unusually powerful hardware.
According to Meta’s announcement, the model has 42 billion active parameters and supports multimodal tasks. The company describes it as its most capable U.S. open-weight model, citing internal benchmark results.
Meta Muse Spark OpenAI Model: Mac Requirements
Meta optimised the model for Apple’s M5 Ultra Mac Studio using the MLX framework. Local deployment requires two Mac Studios with approximately 1TB of combined unified memory.
That hardware requirement makes the model impractical for most consumers. As Engadget reported, the memory demands create a considerable financial barrier. Developers without suitable Macs can instead access the model through supported cloud and API services.
AI Coding and Video Editing Capabilities
Meta demonstrated Muse Spark Open alongside Muse Code, its AI coding terminal. In one example, the system modified an open-source Mastodon client. The tasks included fixing bugs, adding features and building the application.
The company also tested the model with DaVinci Resolve. During the demonstration, it reviewed footage, identified useful moments and cropped video into vertical formats.
These examples highlight Meta’s focus on autonomous workflows. Rather than simply generating responses, the model can interact with software and complete multiple steps. Its performance in independent real-world deployments remains to be established.
Open Weights Give Developers More Flexibility
The model’s open-weight approach allows qualifying developers to download and deploy its parameters. Organisations can customise deployments and maintain greater control over their computing infrastructure.
However, open weights do not mean unrestricted use. Meta distributes the system under its Muse Model License, which includes conditions for commercial applications.
Muse Spark Open is also available through cloud services and API providers. These options let developers experiment without buying expensive workstations.
The Meta Muse Spark Open AI model expands the options for running advanced AI outside traditional cloud infrastructure. Its substantial hardware demands still limit local use to well-equipped developers and organisations.
