QDNASales and integration of LLM inference and training platforms, on-premises or hybrid

Running Kimi K3 on GB300 NVL72

Kimi K3 has 2800 billion parameters. GB300 NVL72 offers 20,700 GB of HBM3e across a 72-GPU rack memory. This page puts the two side by side.

Short answer. Kimi K3 fits comfortably on GB300 NVL72. In FP16, its weights take about 6,440 GB of the 20,700 GB available, leaving 14,260 GB for the attention cache and concurrency.

How much memory does Kimi K3 need?

Weights take the parameter count multiplied by the format size, plus a 15% runtime margin for activations and buffers. The attention cache sits on top: it grows with context length and with the number of concurrent requests, so it is sized case by case.

FormatWeights in memory On this machineWhat the format costs
FP166,440 GBfitsfull precision, the quality reference
FP83,220 GBfitsnegligible loss on most tasks
NVFP41,610 GBfitsBlackwell format, native FP4 compute
Q41,610 GBfitsblock quantisation, broadly supported

Kimi K3 activates 50 billion parameters per token out of 2800 billion. Active parameters govern speed, not footprint: every expert stays resident in memory. Confusing the two under-sizes the machine by an order of magnitude.

What is left to serve requests?

Kimi K3 fits comfortably on GB300 NVL72. In FP16, its weights take about 6,440 GB of the 20,700 GB available, leaving 14,260 GB for the attention cache and concurrency. The remaining margin decides how many concurrent requests and how much context the machine sustains. A margin below a quarter of memory forces a limit on context or on concurrency.

GB300 NVL72 ranges from from 3,9 M€, indicative and not contractual. The machine targets the GE segment.

Where else can Kimi K3 run?

See the Kimi K3 and GB300 NVL72 fact sheets.

How much memory does Kimi K3 need?

In FP16, weights take about 6,440 GB including the runtime margin. The attention cache sits on top and depends on context.

Does Kimi K3 fit on GB300 NVL72?

Kimi K3 fits comfortably on GB300 NVL72. In FP16, its weights take about 6,440 GB of the 20,700 GB available, leaving 14,260 GB for the attention cache and concurrency.

Method

Parameter counts and memory figures come from the site fact sheets. Memory footprints are calculated, not measured: a reading on real hardware may differ depending on the engine and the exact weight format. Prices are indicative and not contractual.