Mac Studio With M3 Ultra Runs Massive DeepSeek R1 AI Model Locally
SMRTR summary
Apple's new Mac Studio with M3 Ultra chip can run large AI models locally when equipped with 512GB of memory. The system efficiently handles a 671 billion parameter DeepSeek R1 model, delivering 17-18 tokens per second while consuming under 200 watts of power. This capability offers privacy advantages for sensitive applications like healthcare data analysis. However, a Mac Studio configured for this performance starts at $10,000, with fully maxed-out versions costing up to $14,099.
SMRTR provides this summary for quick context. The original article belongs to Mac Rumors.
Read the original article