What is Meta Muse Spark Open?

Large language models are machine-learning programs trained on a vast amount of text so they can generate, summarize, translate and analyze language, and they sit behind many modern chatbots. Muse Spark Open is one such model from Meta. Its model card describes it as an open-weights model with 816 billion total and 42 billion active parameters (the adjustable numbers inside the model) and a 1M-token context window, meaning it can take in very long inputs. The card lists a mixture-of-experts design, in which several specialised networks divide up the work, and says the model is built for coding and for agentic work (AI that plans and carries out multi-step tasks) and is natively multimodal (handles more than text, such as images). The card gives its release date as October 15, 2026.

Written from primary sources and reference pages, and checked against them by an independent AI fact-checker. Drafted with AI assistance.