Apple ML Research: Muse Spark Open running on MLX across Mac Studios
Apple's Machine Learning Research site published a technical breakdown of running Meta's 816B-parameter open-weight Muse Spark Open across 2–4 M5 Ultra Mac Studios using MLX, Thunderbolt 5 RDMA, and pipeline/tensor parallelism — 15.7 tok/s (2 Macs, pipeline), 25 tok/s (2 Macs, tensor), 36.7 tok/s (4 Macs, tensor) on the 4-bit model.
Why it matters
Frontier-scale open model running on desktop hardware — the strongest signal yet that local frontier inference is becoming practical, backed by concrete vendor benchmarks rather than marketing claims.
What you could build with it
A privacy-sensitive enterprise (legal, healthcare) could run the model across a small cluster of Mac Studios to keep agentic coding and document work fully on-premises instead of sending data to a cloud API.
Does it hold up?
Evidence is vendor-side (Meta/Apple benchmarks) but unusually concrete with published tokens/sec across setups; independent reproductions will determine real-world practicality given the ~$20k+ hardware cost.
Built with Apple ML Research
- Meta and Apple Bring Muse Spark Open AI Model to Mac StudioMacRumors · Coverage of the Meta-Apple collaboration and on-device agentic demos (Muse Code on Ice Cubes, DaVinci Resolve).
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71JetBrains Mellum 2.1: 12B MoE coding model trained for agentsJetBrains released Mellum 2.1, a 12B MoE model (2.5B active parameters, 128K context) under Apache 2.0, trained mainly…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.