BackNews
General

Meta Releases Muse Spark Open: 81.6B MoE, 1M Context, Apple Partnership

Meta has released Muse Spark Open, an 81.6-billion-parameter mixture-of-experts open-weight AI model with native multimodality and a 1 million token context window. The company claims it is the strongest open-weight model developed in the US, and has partnered with Apple to enable local deployment on Mac computers. The release expands options for creators seeking customizable, locally runnable AI tools for creative workflows.

Meta Releases Muse Spark Open: 81.6B MoE, 1M Context, Apple Partnership
MagicBoat AI4 min read96 views
Table of contents

1. Release Overview

Meta’s new open-weight multimodal model brings high-capacity local AI to Mac users, expanding options for creators who prioritize control and privacy.

On October 16, 2026, Meta unveiled Muse Spark Open, its latest open-weight artificial intelligence model, alongside a partnership with Apple to enable local deployment on Mac computers. The release positions Meta as a major player in the open-weight AI space, with claims that the model outperforms other US-developed open-weight alternatives.

The announcement comes amid a wave of rapid frontier model releases from Google and Meta that have narrowed the performance gap with closed-model leaders like OpenAI and Anthropic, shifting industry competition toward pricing, openness, and ecosystem integration.

2. Core Model Specifications

Muse Spark Open is built on a mixture-of-experts (MoE) architecture with 81.6 billion total parameters. MoE designs activate only a subset of parameters for each input token, making the model more efficient to run than dense models of similar total size without sacrificing performance.

Key confirmed attributes of the model include:

  • Native multimodality, allowing it to process and generate multiple types of content within a single model framework
  • A 1 million token context window, enabling work with long documents, large creative briefs, and extended project files
  • Open-weight licensing, meaning full model weights are available for download, modification, and local deployment by eligible users

Meta has stated that Muse Spark Open is the strongest open-weight model developed in the United States, though specific benchmark results were not included in the initial announcement.

MoE model architecture diagram
MoE model architecture diagram

3. Apple Mac Local Deployment Partnership

A standout detail of the Muse Spark Open launch is Meta’s partnership with Apple to optimize the model for local deployment on Mac devices. Local deployment means the model runs entirely on the user’s hardware, rather than sending data to cloud servers for processing.

This collaboration leverages Apple’s silicon optimization to deliver strong performance on consumer Mac hardware, making large multimodal models accessible to users without enterprise-grade cloud infrastructure. For creators, local execution offers tangible benefits for sensitive projects, offline work, and reduced reliance on stable internet connectivity.

Specific rollout timelines for Mac deployment were not detailed in the initial announcement, but the partnership confirms Meta’s focus on making open models accessible across mainstream consumer devices.

MacBook running local AI model
MacBook running local AI model

4. What This Means for AI Video and Image Creators

While Muse Spark Open’s exact creative use cases will emerge as the community explores the open weights, the model’s core features address several common pain points for AI video and image creators.

Open-weight access gives creators the ability to customize the model for specific creative workflows, from refining script structures to generating detailed reference prompts for text to video pipelines. Local deployment further adds a layer of privacy for creators working on unreleased projects, sensitive client work, or proprietary visual styles.

The 1 million token context window is particularly useful for long-form creative projects, such as feature-length image to video sequences, multi-scene storyboards, or full production scripts. Creators can feed entire project briefs, reference libraries, and style guides into the model in a single pass, reducing the need for fragmented prompting and manual context window management.

For creators who prefer a streamlined, no-setup workflow, multi-model platforms like MagicBoat integrate leading closed and open AI tools in a single interface, eliminating the need to download and optimize models locally while still offering access to cutting-edge creative capabilities. You can try these models for free to see how they fit into your creative process.

AI creative workflow dashboard
AI creative workflow dashboard

5. Broader Landscape Context

The launch of Muse Spark Open is part of a larger shift in the global AI industry, as rapid releases from Google and Meta have ended the absolute frontier performance lead previously held by OpenAI and Anthropic. With performance gaps narrowing across top models, factors like pricing, model openness, and ecosystem integration are becoming the key differentiators for users and developers.

Open-weight models like Muse Spark Open expand the options available to creators, who can now choose between closed, managed models with polished out-of-the-box features and open, customizable models that can be tailored to specific production needs. This growing competition is driving faster innovation, lower costs, and more flexible tools for creative professionals across all visual and digital mediums.

Key Takeaways

  • Meta released Muse Spark Open on October 16, 2026, an 81.6B-param MoE open-weight model with native multimodality and a 1M-token context window.
  • Meta claims Muse Spark Open is the strongest US-developed open-weight model, with a formal partnership with Apple for optimized local Mac deployment.
  • Open-weight access and local execution offer creators greater privacy, customization, and control over their creative workflows and intellectual property.
  • Creators can leverage open models alongside dedicated creative tools via unified platforms to build efficient, flexible production pipelines.

Keywords

muse spark open, meta ai, open-weight ai model, multimodal ai, ai video model, text to video, image to video, free ai tools