Llama 3.1 405B
Apache 2.0
Meta
You chose to run Llama 3.1 405B (405B parameters) — that needs at least 486GB of GPU memory (405B × 1GB × 1.2 working room).
Minimum real build that clears 486GB
1 × DGX H100 (640GB each)
$200,000
Combined memory
640GB
vs. 486GB minimum needed.
Combined power
10,200W
about 8.5 average homes running non-stop.
Total price
$200,000
The whole build, ready to run.
Why this hardware, not something elseWhy one DGX H100 and not, say, 6 separate H100 cards? A single DGX H100 is already 8 H100s pre-wired with NVLink into one 640GB machine — it clears the 486GB minimum in one purchase, one support contract, and one power cable, instead of you having to buy 6+ individual cards and build the fast interconnect between them yourself. For a model this size, buying the packaged system is simpler and just as capable.
Qwen3 235B
Apache 2.0
Alibaba
You chose to run Qwen3 235B (235B parameters) — that needs at least 282GB of GPU memory (235B × 1GB × 1.2 working room).
Minimum real build that clears 282GB
4 × A100 80GB (80GB each)
$48,000
Combined memory
320GB
vs. 282GB minimum needed.
Combined power
1,600W
about 1.3 average homes running non-stop.
Total price
$48,000
The whole build, ready to run.
Why this hardware, not something elseWhy A100 and not the newer H100? Because H100's extra speed mostly helps with training new models and serving many users at once — it doesn't add any GPU memory. Since the only hard requirement here is clearing 282GB, 4 A100s do it for $48,000 and 1,600W, versus roughly double the price and power for the same memory in H100s. We only recommend the faster (pricier) card when the workload actually needs the speed.
DeepSeek-V3 671B
MIT
DeepSeek
You chose to run DeepSeek-V3 671B (671B parameters) — that needs at least 805.2GB of GPU memory (671B × 1GB × 1.2 working room).
Minimum real build that clears 805.2GB
2 × DGX H100 (640GB each)
$400,000
Combined memory
1,280GB
vs. 805.2GB minimum needed.
Combined power
20,400W
about 17 average homes running non-stop.
Total price
$400,000
The whole build, ready to run.
Why this hardware, not something elseWhy 2 DGX H100 systems? One alone (640GB) is short of the 805GB this model needs — it would fail to load, full stop. Two is the smallest number of complete, real-cluster systems that clears the bar (1,280GB). We didn't recommend the GB200 NVL72 rack here: it would also work (13,500GB), but at $6.5M for a single rack it's roughly 16x the price for far more capacity than this one model needs — the honest minimum is 2 DGX H100 systems.