ARTIFICE DAILY AI + CYBER SECURITY BRIEFING

Artificial Intelligence

Meta releases Muse Spark Open for Apple M5 Ultra Mac Studios

Meta AI Research released Muse Spark Open, an open-weight multimodal model, and said it could run locally on two Apple M5 Ultra Mac Studios using the MLX framework.

Meta AI Research

Meta AI Research said it had released the weights for Muse Spark Open. It described the model as multimodal, with 816B total parameters, 42B active parameters and a 1M-token context window.

The company said the model was built for coding and agentic work. It said the model could run locally on two Apple M5 Ultra Mac Studios with the latest MLX framework, and that Apple's memory design meant the two machines could hold the model together.

Meta said the weights were available through API platforms and download services listed in the post. It also said it had tested on-device workflows that included changes to Ice Cubes and a full run through DaVinci Resolve.

The post said Meta had evaluated the model under its Meta Superintelligence Scaling Framework. It said the release used the Muse Model License, which carried limited commercial terms.

Named in this story

Companies

Apple
made the M5 Ultra Mac Studios Meta said it used

Organisations

Meta AI Research
published the release note for Muse Spark Open

Products and systems

Muse Spark Open
the open-weight model Meta released
Apple M5 Ultra Mac Studios
the desktop machines Meta said could run the model locally
MLX framework
the framework Meta said enabled the local run
Ice Cubes
the open-source Mastodon client used in Meta's test workflow
DaVinci Resolve
the video editor Meta said the model drove end to end
Meta Superintelligence Scaling Framework
Meta's framework for release evaluation
Muse Model License
the licence Meta used for the release

How the source tells it

The piece reads like a launch post, with vendor boosterism and novelty claims carried by superlatives and first-in-class framing.

  • vendor boosterism the release is built around repeated self-praise about capability, access and performance
  • novelty it relies on first-in-class language and launch-day availability to present the model as new