Latest
Category 2 Hurricane Isaias nears landfall, leading to production shut-insAmazon devices chief Panos Panay on the importance of failure and bouncing back from itFamily Separation Causes Real and Lasting Harm for Children and ParentsMitch Daniels rebukes JD Vance over free-market retreatTrump, after bipartisan backlash, says Pentagon won't livestream execution of Fort Hood shooterZacharia names Scarlett Johansson as actress she wants to play her on ‘SNL’FBI investigates letters with ‘unknown substance’ sent to US election officialsCategory 2 Hurricane Isaias nears landfall, leading to production shut-insAmazon devices chief Panos Panay on the importance of failure and bouncing back from itFamily Separation Causes Real and Lasting Harm for Children and ParentsMitch Daniels rebukes JD Vance over free-market retreatTrump, after bipartisan backlash, says Pentagon won't livestream execution of Fort Hood shooterZacharia names Scarlett Johansson as actress she wants to play her on ‘SNL’FBI investigates letters with ‘unknown substance’ sent to US election officials
← Signal: Tech & AI

Meta releases Muse Spark Open, an open-weight AI model optimized to run locally on Mac computers

Confirmed1 source · Oct 15, 2026

The 816-billion-parameter model is designed for coding and complex tasks and requires at least 1 terabyte of RAM to run locally.

Meta releases Muse Spark Open, an open-weight AI model optimized to run locally on Mac computers
Image via TechCrunch

What happened

Meta launched Muse Spark Open, which it describes as the most capable open-weight model available in the United States. The model features 816 billion parameters, a 1 million token context window, and is natively multimodal. It has been specifically designed to run locally on Mac computers, particularly the Mac Studio equipped with the M5 Ultra chip, though it requires at minimum 1 terabyte of RAM. Meta paired the model with Muse Code, its AI coding terminal, and deployed it to various workflows including bug fixes and new features for a Mastodon client, as well as video editing tasks using DaVinci Resolve. The model is now available through multiple API platforms.

Context

Open-weight models offer cost efficiency, transparency, and flexibility advantages over closed proprietary systems, which positions this release as a potential shift in how developers access high-performance AI locally rather than relying on cloud-based alternatives. The focus on Mac compatibility and complex coding tasks suggests Meta is targeting professional developers and knowledge workers who work on Apple hardware. However, the extremely high hardware requirements—1 terabyte of RAM and specialized Mac processors—constrain the addressable market to users with access to high-end computing resources.