QDNAAdvisory and architecture for LLM inference and training platforms, on-premises or hybrid

Running Kimi K3 on B300 SXM

Kimi K3 has 2800 billion parameters. B300 SXM offers 2,304 GB of HBM3e across 8 GPUs memory. This page puts the two side by side.

Short answer. Kimi K3 fits on B300 SXM only when quantised. In NVFP4, its weights take about 1,610 GB of 2,304 GB. More precise formats do not fit.

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. In NVFP4 the calculation reads: 2,800 billion parameters × 0.5 byte (NVFP4) = 1,400.0 GB of weights; × 1.15 runtime margin = 1,610.0 GB. 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 GBdoes not fitfull precision, the quality reference
FP83,220 GBdoes not fitnegligible loss on most tasks
NVFP41,610 GBfitsBlackwell format, native FP4 compute
Q41,932 GBtightblock quantisation, broadly supported

Model sheet: 2,800 billion parameters, 104 billion active per token, 1,048,576-token context, Kimi K3 licence. 2.8 trillion announced, 16 of 896 experts active; the repository holds 2,780 billion elements, published in MXFP4 (1,561 GB).

Kimi K3 activates 104 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 on B300 SXM only when quantised. In NVFP4, its weights take about 1,610 GB of 2,304 GB. More precise formats do not fit. The remaining margin decides how many concurrent requests and how much context the machine sustains. By QDNA convention a machine “fits” when the weights take at most 75% of its memory; beyond that it “barely fits” and forces a limit on context or on concurrency. The attention cache is not calculated on this page. These models’ architectures (compressed latent, linear or sparse attention, Mamba layers) share no per-token formula; it is measured on the machine, with the target context and concurrency [TO BE MEASURED].

B300 SXM: indicative price about €430k to €690k excl. VAT, not contractual. Source: BIZON X9000 G5 at $496,728 (November 2026 delivery) and Supermicro AS-8126GS-NB3RT at $795,000 from Vipera, both listed on 2 September 2026, converted at the ECB rate of $1.159 per €1. The machine targets the GE segment.

Where else can Kimi K3 run?

See the Kimi K3 and B300 SXM fact sheets.

How much memory does Kimi K3 need?

In NVFP4, weights take about 1,610 GB including the runtime margin. The attention cache sits on top and depends on context.

Does Kimi K3 fit on B300 SXM?

Kimi K3 fits on B300 SXM only when quantised. In NVFP4, its weights take about 1,610 GB of 2,304 GB. More precise formats do not fit.

Method and sources

Memory footprints are calculated, not measured: a reading on real hardware may differ depending on the engine and the exact weight format. Every figure below carries its source and the date of its last check.