What the AI world thinks about Meta Muse Spark Open
On Oct 15 Meta dropped Muse Spark Open, an 816B open-weight coding/agent model you can apparently run on two Mac Studios. Another open-weight race, another week of being told America is back. Here is what the roster said, copied exactly.
Today we’re releasing Muse Spark Open, the most capable US open-weight model to date. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning. Key Model Specs: → 816B total / 42B active parameters → 1M token https://t.co/yAtoxNeYJm [excerpt: post continues on X]
America is SO back Zuck just announced Muse Spark Open is dropping later today, built with Mac Studios in mind This is bigger than you think The Meta Muse model was trained specifically for agent tasks. Now we can run it locally, plugged into agents like Hermes This is the https://t.co/1DU5al6kJz [excerpt: post continues on X]
🚨 Meta just open-sourced Muse Spark!! - best US open-source - better than Kimi and GLM - multi-modal - perfect for custom fine-tunes BRILLIANT move from Meta as it races to catch up with the other AI labs https://t.co/F7n7NJw3dU [excerpt: post continues on X]
And @AIatMeta went live with Muse Spark Open! The 816B model was trained on NVIDIA GB300 NVL72. Its 1M-token context window gives coding agents room to work across large codebases. Meta also produced NVFP4 checkpoints optimized for Blackwell Ultra, delivering over 50K https://t.co/HJUKdC1n3D [excerpt: post continues on X]
llama serve -hf ggml-org/Muse-Spark-Open-816B-A42B-GGUF:Q2_K (on two mac studios). Well done @finkd & @johnternus! https://t.co/Kjl7qbPV52
@HarryStebbings da fuck you talking about? we got so many models on our API. you want spark, we got spark. you want wonder, we got wonder. you want image, we got image. https://t.co/jh22Ufz45h
lmao muse spark open balling out https://t.co/AcUT9o8NEQ
Meta just released Muse Spark Open, calling it "the most capable US open-weight model to date." Meta puts it at 39 on Artificial Analysis’ Intelligence Index, leading U.S. rivals but still behind China’s GLM 5.3 (45) and Kimi K3 (44). Nvidia also shipped its own open model https://t.co/yTB2jMJMF0 [excerpt: post continues on X]
To help make advanced models more accessible, Muse Spark Open is optimized to run on two Apple M5 Ultra Mac Studios with the latest MLX framework. It also runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, llama.cpp, and Unsloth with [excerpt: post continues on X]
Muse Spark Open is available today through API platforms at Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI. Download the weights on Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face, and Microsoft Foundry Models. Weights are released under the [excerpt: post continues on X]
@AIatMeta Yesss 👏👏👏
@finkd @johnternus Awesome 👍👏👌
Get the roundup, skip the discourse
One short email when something actually big happens. No threads, no hustle, no “🧵 1/47”.