Llama 3.1 405B Instruct
Llama 3.1 405B Instruct loads in 243 GB at Q4_K_M and is comfortable from 256 GB of memory. Below: the full memory math, the cheapest card that runs it well, and the fastest machine we track.
PARAMETERS
405B
FORMAT
Q4_K_M
MIN MEMORY
320 GB
BEST FOR
Quality, Reasoning, Coding
Memory math
Weights (Q4_K_M)
243 GB
+ KV cache (16k)
~6.0 GB
Total at 16k
~249.0 GB
Comfortable from
256 GB
KV = fp16 estimate (q8_0 cache roughly halves it). "Comfortable" = weights + KV within the engine's tiered budget (~70-85% of memory).
Hardware snapshot
Cheapest GPU
—
no consumer fit
Fastest
—
n/a
Runs on a Mac?
No
beyond current Macs
More Llama models
Frequently asked questions
How much memory does Llama 3.1 405B Instruct need?
243 GB for the Q4_K_M weights, plus ~6.0 GB of KV-cache at 16k context — about 249.0 GB total. Comfortable from 256 GB of VRAM or unified memory.
Does Llama 3.1 405B Instruct run on a Mac?
Not on the current Mac configs we track; Llama 3.1 405B Instruct needs more memory than they offer.
What is the cheapest GPU for Llama 3.1 405B Instruct?
No tracked consumer GPU runs Llama 3.1 405B Instruct comfortably.
Cite this page
ModelFit: Llama 3.1 405B Instruct — specs, memory math and hardware verdicts. https://modelfit.io/models/llama-3.1-405b/ (dataset updated 2026-09-03, CC BY 4.0).