Apple and Meta Partner on Local AI Model Deployment

Apple and Meta have announced a partnership that will allow developers to run Meta's Muse Spark Open AI model locally on Apple's high-end Mac computers. This collaboration signifies a growing industry interest in processing complex AI models on local hardware rather than relying solely on cloud infrastructure, enhancing security and potentially reducing costs.

The partnership is notable given the historical rivalry between the two tech giants. The Muse Spark Open model, featuring an 816-billion-parameter "mixture-of-experts" architecture, is an open-weight model. This allows developers to download its parameters and run it on various hardware configurations. Users can reportedly combine multiple Mac Studio computers to process tasks like coding and video analysis.

Meta's head of AI, Alexandr Wang, stated that while many powerful open-weight models are too large for typical personal computers, Apple's Mac Studio provides sufficient memory to make running Muse Spark Open practical. This move could present a challenge to other major AI players like OpenAI and Anthropic, while also highlighting Apple's advancements in its own AI chip technology.

Apple has also released new benchmarks for its chips and technical documentation for its MLX AI framework, which supports splitting open-source models across multiple machines. However, the solution requires high-end Mac Studio machines, with a base configuration meeting the necessary specifications costing upwards of $10,799.