Models Published:
Meta releases Muse Spark Open, an 816B open-weight model it runs on two Mac Studios
What happened
Meta released Muse Spark Open on October 15, an open-weight, natively multimodal model with 816B total parameters, 42B of them active, and a 1M-token context window, aimed at coding and agentic work. The weights are on Hugging Face, where access must be requested, together with an NVFP4 version, under the Muse Model License, which has limited commercial terms. Meta says the model runs on SGLang, vLLM, TensorRT-LLM, Transformers, Apple MLX, llama.cpp and Unsloth, and it demonstrated it on two Mac Studios with M5 Ultra, each with up to 512 GB of unified memory. It is also offered through hosted APIs such as Groq, Fireworks AI, Together AI and OpenRouter.
Why it matters locally: A frontier-size open model from a major US lab now has an official path to run on desk-side hardware, not only in a data center.
Three ways to read it
For the market
Meta calls it the most capable US open-weight model to date, and making it available on Hugging Face, three big clouds and seven API providers at once shows it wants wide adoption despite the limited licence. Showcasing Apple hardware rather than only GPUs likely strengthens the Mac Studio's case as a quiet alternative to a GPU server for very large models.
In plain words
Meta has handed out one of its biggest AI models so anyone with the right machines can run it at home or in the office. It is still huge: Meta's own demo needed two top-spec Mac Studios working together.
For practitioners
Only 42B parameters are active per token, so speed depends more on memory bandwidth than on total size, but the full 816B must still fit in memory; plan for two 512 GB M5 Ultra Mac Studios with MLX, or a multi-GPU server with the NVFP4 weights. Apple says the 512 GB Mac Studio configuration ships in late October 2026. Read the Muse Model License before any commercial deployment, and request Hugging Face access early because the repos are gated.
Sprout's insight
Sprout compares this with other brands and with where real applications are heading. Free for members.
Sign in free to read it →Subscribe to Sprout's weekly
One email a week: the top stories and a taste of Sprout’s insights. Unsubscribe any time.
Source
Read the official source: Introducing Muse Spark Open (Meta AI) ↗Summaries are written by OnPrem World from the official source. We link to the original and never republish it.
On this site
More news
- vLLM v0.32 brings 4-bit KV cache to more GPUs and stops allowing every CORS origin by default
- DeepSeek-V4.1-Flash-1017: a same-size update with better agent scores and shorter answers
- Open WebUI 0.12.3 adds end-to-end encrypted chats and one search box for the whole workspace
- llama.cpp v0.7.0 adds a GPU cache for offloaded MoE experts and moves the server to port 9931
- Mistral releases Timestral, time-series forecasting models from 5M to 209M parameters that run on a CPU