Friday, October 23, 2026
Herald of AI

AIGenerated AI news for everyone, written by AI and explained simply. Models, technology, culture and the people behind it.

New models

Meta launches Muse Spark Open, an open-weight model built to run on Macs

The new 816-billion-parameter model can run locally on Apple computers, but you will need serious hardware to make it work.

AIGenerated · Written by AI from the sources listed below · Read and approved by a person before publication
Close-up of hands typing on a laptop keyboard, ideal for tech and work scenes.
Close-up of hands typing on a laptop keyboard, ideal for tech and work scenes. File picturePhoto: Atlantic Ambience / Pexels

A new open model for Apple computers

Meta has released a new open-weight AI model called Muse Spark Open, claiming it is the most capable open-weight model available in the United States. The model was designed from the ground up to run locally on Apple computers, including the Mac Studio equipped with the M5 Ultra chip.

The model uses 816 billion parameters and can handle up to 1 million tokens of context, which means it can process very long documents or conversations in one go. Meta says it is natively multimodal and built for complex coding work and agentic tasks, which are AI systems that can carry out multi-step workflows on their own.

Not exactly portable

There is a catch. Running Muse Spark Open locally requires at least 1 terabyte of RAM, a specification that currently limits the model to high-end desktop computers rather than laptops. The Mac Studio with the M5 Ultra chip is recommended as the best option for running the model locally.

For most people without that much memory available, Meta is making the model accessible through several API platforms instead of requiring a local installation.

Real-world tasks

Meta demonstrated the model by pairing it with Muse Code, its own AI coding terminal, and putting it to work on a coding project for a Mastodon client. The company said the model handled bug fixes, added new features, and completed a full build of the software. It was also tested with DaVinci Resolve, a professional video editing program, where it reviewed and cropped footage successfully.

These examples show how open-weight models like Muse Spark Open aim to handle real production tasks, not just chat or simple questions.

Why open weight matters

Open-weight models give developers more flexibility than closed or proprietary models. The source code and weights are publicly available, which means companies can run the models on their own servers and avoid paying for cloud API access. This can reduce costs significantly over time and give teams more control over how the model behaves in their systems.

According to benchmark data from Artificial Analysis, Muse Spark 1.3, which is closely related to the new release, holds an intelligence index of 48.1, placing it in a competitive range among frontier-class open models.

Meta says the model will be available now through multiple API platforms, giving developers and researchers immediate access to test it out before deciding whether to run it locally on their own hardware.

Key facts

  • Muse Spark Open has 816 billion parameters and a 1 million token context window
  • Running the model locally requires at least 1 terabyte of RAM
  • Meta designed the model to run on Apple computers, especially the Mac Studio with M5 Ultra
  • The model was tested on coding tasks for Mastodon and video editing in DaVinci Resolve
  • Open-weight models let developers run AI locally without paying for API access

Sources

  1. Meta launches an open-weight AI model that runs on Macs | TechCrunch — techcrunch.com
  2. AI boom — en.wikipedia.org

This story was written before each paragraph carried a source number: every fact in it comes from one of the sources above.

Related stories