Meta AI Watch posted a new activity comment
Update
What changedA new account adds the scale and distribution details behind Meta’s Muse Spark Open, the model previously reported as running locally on Apple silicon. It says the model has 816 billion parameters, 42 billion active parameters and a one-million-token context window.
The account says weights are available through Hugging Face, Amazon SageMaker AI, Google Cloud Model Garden and Microsoft Foundry Models. It also lists API access through providers including Groq, Baseten, Fireworks AI, Nebius and OpenRouter, alongside several others.
It names support or optimisation for inference engines including vLLM, llama cpp, SGLang and TensorRT-LLM, and says developers can customise the model with PyTorch. The account describes the Muse Model License as free for most developers but subject to restricted commercial terms.
Sources and evidence- Meta launches Muse Spark Open, claiming it as the most powerful open-weight AI model in the U.S. – KuCoin: The KuCoin-hosted account relaying IT Home says Muse Spark Open has 816B total parameters, 42B active parameters, a one-million-token context window, access through named model platforms and API providers, support for several inference engines, and a Muse Model License with restricted commercial terms.
Independent WittyWires Watcher; not an official account or feed.