Apple and Meta enable 816-billion-parameter AI model on Mac Studio desktops
AI
Dispatched by Tariq Al-Mansoor, Gulf & Capital Desk
- Meta’s 816-billion-parameter Muse Spark Open can run on two wired-together Mac Studio desktops.
- The setup lets companies process prompts locally, avoiding token-based cloud fees and keeping data in-house.
- Enterprises such as PNC, Siemens and fintech firms are evaluating the approach to cut AI operating costs.
Apple and Meta have tuned the Muse Spark Open model to run on paired Mac Studio machines, offering on-premise AI for businesses.
«Apple, Meta Team Up to Put Big AI on Desktops»
Summary and commentary by Aura-V. Facts, figures and opinions belong to the original.
Meta released Muse Spark Open, an 816-billion-parameter model it calls the most capable U.S. open-weight model, and announced a joint effort with Apple to tune it for two Mac Studio desktops.
Apple demonstrated that two Mac Studios wired together can host Muse Spark Open at a usable speed, with four machines processing faster than a typical reading pace. The collaboration gives Meta a hardware showcase outside its data centers and offers Apple a privacy-focused AI story, as the model runs locally and never touches user data.
Businesses facing high AI cloud bills—including legal AI firm Harvey, healthcare startup Abridge, fintech Ramp and AT&T—are looking at on-premise models to reduce costs. PNC Financial Services, CH Robinson, Siemens and Tinder are also considering open-weight models to avoid rising cloud expenses, according to PYMNTS. Running large models on Mac Studio promises “no token counting” and keeps customer data within the organization.
Interested in implementing this technology inside your company?
Our studio designs custom agent meshes, data pipelines, and outbound GTM systems for forward-looking enterprises.
