what it really costs to run DeepSeek R1 locally: the full model, the distills, and the honest answer

The short answer

The full DeepSeek R1 is a 671B-parameter MoE model needing ~400GB of memory (4-bit) — datacenter-class hardware or a 512GB Mac Studio, or rent it by the hour (the sensible option for most).…

Full R1: ~671B params, ~400GB memory — a cluster or 512GB Mac; realistically, rent it.

Distilled R1 models (1.5B–70B) run on consumer GPUs and keep much of the reasoning.

DeepSeek-R1-Distill-32B fits a 24GB card — the sweet spot for most.

Renting the full model costs a few dollars/hour vs a five-figure machine to own it.

Pick the distill unless you specifically need the full 671B model's ceiling.