Meta has an artificial intelligence model that needs 1 TB of memory to operate
Meta has released Muse Spark Open, an 800-billion parameter model designed for local deployment on Apple M5 Ultra-based hardware. The model requires 1 TB of unified memory, achievable by clustering multiple Mac Studio u
Meta has released Muse Spark Open, an 800-billion parameter model designed for local deployment on Apple M5 Ultra-based hardware.
The model requires 1 TB of unified memory, achievable by clustering multiple Mac Studio units.
Documentation suggests it outperforms current benchmarks like GLM 5.3 and Kimi K3 in complex reasoning and visual tasks.
The model is now available on Hugging Face.
The model has 800 billion parameters. Meta describes it as the most capable artificial intelligence model in the US to date, among those whose files defining its operation are available for download.
In tests published by the company, Muse Spark Open outperforms GLM 5.3 and Kimi K3 on tasks that require it to complete complex steps and draw conclusions from images. These are Meta test results, not an independent comparison.
Muse Spark Open is available for download from Hugging Face. It can also run on four Mac Studios with 256GB of memory each. According to the report, these will cost a total of about $38.000.
READ ALSO
Solidigm seeks US permission to ship chip equipment to China in exchange for new factory
Apple hasn't announced a specific price or release date for the 512GB Mac Studio. It has said it will be available in late October. Meta developed the model in collaboration with Apple and John Ternus.
Meta says the model may be of interest to large organizations that spend a lot on using AI models, as long as it performs as it claims.
G
Hacks.gr on Google
Set it as a preferred source for cybersecurity updates.
Preferred source
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.