Editorial illustration for Meta turns Mac into AI supercomputer; gender gaps show in training, not access
Meta's 816B Muse Spark Model Challenges OpenAI
Meta turns Mac into AI supercomputer; gender gaps show in training, not access
Meta spent the week proving it can still ship. Monday brought Muse Wonder, the frontier model meant to close the gap with OpenAI and Anthropic. Now comes Muse Spark Open, an 816-billion-parameter multimodal model that Meta is calling the most capable open-weight release to come out of the U.S.
The bigger surprise is who helped build the on-ramp: Apple. Meta tuned Spark Open to run on two M5 Ultra Mac Studios wired together through Apple's updated MLX toolkit, turning a pair of desktop machines into something closer to a small AI supercomputer. Meta says the setup handled real work, iOS app fixes, video edits in DaVinci Resolve, without touching the cloud.
The timing matters. Artificial Analysis already has Spark Open ranked above every other Western open-weight model on its Intelligence Index, though still behind China's GLM 5.3, Kimi K3, and Qwen3.8. Below, we get into where Spark Open actually wins and loses against those rivals, plus a Harvard report that looks past the usual access questions to ask where AI's gender gap really shows up.
Two M5 Ultra Mac Studios with 1TB combined memory can run Spark Open via Apple’s updated MLX toolkit, with Meta saying the model shipped iOS app fixes and edited video in DaVinci Resolve without the cloud.
Why this matters
Meta pairing Muse Spark Open with Apple's hardware is a bigger deal than another model release. Running frontier-grade AI across two Macs instead of a data center rack changes who can afford to experiment, which matters to any founder or researcher who's been priced out of GPU clusters. That's the access story, and on paper it's genuinely good news.
The gender data complicates the celebration. If usage numbers look equal but training, confidence, and visibility don't, then "AI for everyone" is a slogan that's outrunning the product. We'd push back on any lab framing open access as the finish line.
For developers building teams and researchers citing adoption stats, the numbers worth watching aren't just downloads or API calls. Who's getting credited, promoted, and listened to inside these tools matters just as much.
Muse Spark Open might lower the hardware barrier. Whether it lowers every other barrier is a separate question, and right now the evidence says no. Watch how Meta and Apple talk about who's actually building on this, not just who's running it.
Related Reading
Browse all Meta news →
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source Box CEO Aaron Levie Says AI Will Drive "Biggest Cybersecurity Upgrade
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
Related Topic Meta’s Muse Tracks Your Email and Bank Data While You Work
AI Tools & Apps
Related Topic Meta Faces Lawsuit Over AI Training Data, Facial Recognition
Policy & Regulation
From Archives DeepSeek releases V4.1-Flash-1017, scores 77.0 on DeepSWE, opens Search API
From Archives Mathematicians Criticize OpenAI's Navier-Stokes Proof Release
From Archives Reflection's 501B Beam model is free on Ollama for a limited time
LLMs & Generative AI
From Archives Meta Muse and Apple HomeView surface US AI support for future viewpoints
Research & Benchmarks
Common Questions Answered
What is Muse Spark Open and how does it compare to other frontier AI models?
Muse Spark Open is Meta's 816-billion-parameter multimodal model designed to close the gap with competitors like OpenAI and Anthropic. It represents Meta's most capable open-weight release from the U.S. and demonstrates Meta's continued ability to ship competitive frontier-grade AI models.
How does Meta's partnership with Apple enable running Muse Spark Open on Mac hardware?
Meta optimized Muse Spark Open to run on two M5 Ultra Mac Studios connected through Apple's updated MLX toolkit, with 1TB combined memory. This setup allows the model to perform tasks like shipping iOS app fixes and editing video in DaVinci Resolve without requiring cloud infrastructure.
Why does democratizing AI access through Mac hardware matter for founders and researchers?
Running frontier-grade AI on two desktop Macs instead of expensive data center GPU clusters significantly reduces the barrier to entry for experimentation. This accessibility matters because it allows founders and researchers who have been priced out of traditional GPU clusters to now afford and access cutting-edge AI capabilities.
What gender gap issue does the article identify in AI development despite equal access?
The article notes that while usage numbers may look equal, there are disparities in training, confidence, and visibility between genders in AI. This suggests that simply providing equal access to AI tools does not automatically solve underlying gender representation and confidence gaps in AI development and deployment.
Further Reading
- Introducing Muse Spark Open, The Most Capable US Open-Weight Model to Date - Meta AI
- Meta Makes Its Open Models Comeback - The Deep View
- Meta and Apple Bring Muse Spark Open AI Model to Mac Studio - MacRumors
- Meta Transforms Two Mac Studios into AI Supercomputer - ForkLog
- Meta Unveils 816B-Parameter Muse Spark Open With Apple - The Chosun Daily
