DEVOURED
Meta unveils Muse Spark Open, the new top US open weights model — and it runs on two paired Mac desktops
Meta released Muse Spark Open, an 816-billion-parameter model that runs on dual Mac desktops.
What: The model features a mixture-of-experts architecture activating 42 billion parameters per token and supports a 1-million-token context window with multimodal inputs.
Why it matters: The ability to run massive open-weights models on consumer hardware significantly lowers the barrier to entry for developing and testing frontier-scale AI systems.
Decoder
- Mixture-of-experts (MoE): An architecture where only a subset of the model's parameters are activated for each input, improving computational efficiency.
- Open weights: A distribution model where the learned parameters of a neural network are publicly available, though the training data and code may remain closed.
Original article
Muse Spark Open is an 816-billion-parameter open-weight model that can run on relatively affordable consumer hardware. Meta's benchmark tests used a pair of M5 Ultra Mac Studios with an unknown memory configuration. The model uses a mixture-of-experts design and activates 42 billion parameters for each token. It supports a context of up to 1 million tokens and accepts text, images, video, and audio, but only outputs text.
Explore all 51 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi