News for hackers. Stuff that matters.

A comment on Apple M3 Ultra. Parent comment

Re:Apple M3 Ultra by NightlyDev on Wednesday March 05, 2025 @10:02AM (#43267455)
The full deepseek R1 model needs more memory than 512GB. The model is 720GB alone. You can run a quantized version on it, but not the full model.
[ Parent | on HN ]
Re:Apple M3 Ultra by summarity on Wednesday March 05, 2025 @06:35PM (#43274283)
You can chain multiple Mac Studios using exo for inference, you'd "only" need two of these. There's a bottleneck in the RMA speed over TB5, but this may not matter as much for a MoE model.
[ Parent | on HN ]