AI News · Models ·
Meta releases Muse Spark Open, an 816 billion parameter open-weight model
Meta has launched Muse Spark Open, an open-weight model with 816 billion parameters and a 1 million token context window. The company says it outperforms other popular open-weight models and is aimed at coding and agentic work. It is available now through several API platforms and can run locally on high-memory Macs.
Key points
- Muse Spark Open has 816 billion parameters and a 1 million token context window.
- Running it locally requires at least 1 terabyte of RAM, according to Engadget.
- Meta paired the model with Muse Code, its AI coding terminal.
What happened: Meta launched Muse Spark Open on Thursday, an open-weight model that TechCrunch described as the most capable open-weight model in the United States. The model has 816 billion parameters and a 1 million token context window, and Meta said it is natively multimodal and designed for complex coding and agentic tasks.
The details: Meta released evaluation results it says show the model outperforming other popular open-weight models, though TechCrunch did not report specific benchmark scores. Running the model locally requires at least 1 terabyte of RAM, according to Engadget, and TechCrunch reported it runs best on a Mac Studio with the M5 Ultra chip. Meta paired it with Muse Code, its AI coding terminal, and showed it fixing a Mastodon client, with Meta describing "bug fixes, new features, [and] full build." The company also said the model could review and crop footage in DaVinci Resolve.
Background: The release follows Meta's Muse Wonder frontier models earlier this week. Muse Spark Open is available now through several API platforms, which were not named in the report. License terms and pricing were not reported.
What to watch: Open-weight models can be more cost efficient and offer more transparency and flexibility than closed models, TechCrunch noted. The hardware bar is high, so most teams will likely reach the model through hosted APIs first. Independent benchmark results will show whether Meta's performance claims hold up against leading closed and open models.
Our take
Teams that want a capable model they can run on their own hardware now have a large new open-weight option to evaluate against closed APIs.