DeepSeek-R1 671B
DeepSeek-R1 671B loads in 380 GB at Q4_K_M and decodes at the speed of its 37B active path, comfortably from 256 GB of memory. Below is the memory math we ran for it, the cheapest card that fits, and the fastest machine we track.
How much memory does DeepSeek-R1 671B need?
DeepSeek-R1 671B is a 671B mixture-of-experts model (37B active per token) built for reasoning, coding. The weights alone take 380 GB at Q4_K_M; add context and the OS, and it belongs in the 256 GB memory class. Every number below comes from the ModelFit engine — speeds are estimates, labeled est., because we do not benchmark hardware ourselves.
KV = fp16 estimate (q8_0 cache roughly halves it). "Comfortable" = weights + KV within the engine's tiered budget (~70-85% of memory).
Which machines run DeepSeek-R1 671B?
Verdicts come from the same engine that powers our fit checker — memory capacity first, bandwidth for speed. How we compute these numbers · Hugging Face model card
More DeepSeek models
Frequently asked questions
How much memory does DeepSeek-R1 671B need?
380 GB for the Q4_K_M weights, plus ~6.0 GB of KV-cache at 16k context — about 386.0 GB total. Comfortable from 256 GB of VRAM or unified memory.
Does DeepSeek-R1 671B run on a Mac?
Not on the current Mac configs we track; DeepSeek-R1 671B needs more memory than they offer.
What is the cheapest GPU for DeepSeek-R1 671B?
No tracked consumer GPU runs DeepSeek-R1 671B comfortably.
Cite this page
ModelFit: DeepSeek-R1 671B — specs, memory math and hardware verdicts. https://modelfit.io/models/deepseek-r1-671b/ (dataset updated 2026-09-03, CC BY 4.0).