All videos
0:00 / 0:00
ai

Run an 816B AI model on just two Mac Studios. The future of local compute is here.

Julian Goldie Agency4 November 2026Watch on YouTube

Part of series

Ep. 5 · Meta's 816B model lokaal draaien

Meta's open-weight model Muse Spark Open met 816 miljard parameters draait verrassend lokaal op twee Mac Studios.

View the series

What you'll learn

  • A 816B parameter model from Meta, Muse Spark, can run locally on two Mac Studios with M5 Ultra chips.
  • The unified memory architecture of the M5 Ultra makes it possible to run a model of this size on a desktop setup.
  • The setup works as a personal supercomputer for complex tasks, without the need for large server farms.
  • Several engines such as VLM and SGLang are supported for local inference.
  • The video highlights a shift from cloud dependency to local ownership for developers.

Frequently asked questions

Which model and hardware are used in the video?
The video covers Meta's Muse Spark, an open model with 816B parameters, running on two Mac Studios with M5 Ultra chips.
Why does the unified memory architecture of the M5 Ultra matter here?
According to the video, that architecture makes it possible to run a 816B parameter model on a desktop setup instead of on large server farms.
Which engines for local inference are mentioned?
The video mentions VLM and SGLang as engines supported in this configuration.
Which shift for developers does the video highlight?
The video highlights the move from cloud dependency to local ownership, where developers can run high-end models without renting cloud infrastructure.

Topics

Read next

Description from the channel

Want to make money and save time with AI? Join here: https://www.skool.com/ai-profit-lab-7462/about Meta's new Muse Spark Open model is changing the game for local AI. By leveraging the unified memory architecture of the M5 Ultra chip, it is now possible to run an massive 816B parameter model on a desktop setup rather than needing massive server farms. This configuration creates a personal supercomputer capable of handling complex tasks at incredible speeds, supporting various engines like VLM and SGLang. This shifts the paradigm from cloud-dependency to local ownership, offering a powerful alternative for developers looking to run high-end models without renting cloud infrastructure. #AI #AppleMacStudio #MetaAI #TechNews #LocalLLM