Meta Muse Spark Open: 816B open-weight model for coding and agents
Meta's 816B-parameter open-weight MoE model (42B active, 1M-token context) for coding and agentic work, released under the Muse Model License.
Why it matters
Meta calls it the most capable US open-weight model to date — and co-engineered it with Apple to run on two M5 Ultra Mac Studios, putting frontier-scale inference on a desk.
What you could build with it
A small software team could run the full model across two Mac Studios to power an in-house coding agent with a 1M-token view of their entire codebase, keeping proprietary code off third-party clouds. The result: a private pair-programmer for long-horizon refactors and bug hunts.
Does it hold up?
Too early to judge — weights landed Oct 15 with day-one support across vLLM, SGLang, TensorRT-LLM, MLX, and Unsloth, but no independent hands-on reports exist yet beyond the announcement demos.
Built with Meta Muse Spark Open
- Unsloth: run Muse Spark Open locally (GGUF guide)unsloth · Quantized GGUF release down to ~2-bit with a step-by-step llama.cpp run guide for 256GB machines, published the same day as the weights.
- Meta Muse Spark Open: 816B open-weight AI runs on two Mac Studiosmadrobot.blog · In-depth walkthrough of the Apple co-engineering, the memory math (1.6TB BF16 vs ~493GB NVFP4), and measured token throughput on M5 Ultra Mac Studios.
- Meta Opens the Weights on Muse Sparksalhi.com · Ecosystem roundup: GB300 NVFP4 checkpoints at 600 tok/s, seven API hosts (Baseten, Fireworks, Groq, Modal, Nebius, OpenRouter, Together), and Helion kernel speedups.
Learn more
First spotted on hf: 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.