// RUN โ€” FITS

Can the NVIDIA H100 PCIe 80GB run Generative BST?

Yes โ€” here's how fast and up to what context, with real measured numbers.

Generative BST on the NVIDIA H100 PCIe 80GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 406k 77.4-92.2 t/s ~1.5-1.8 (slow)
Q8 yes 478k 154.8-184.5 t/s ~3.1-3.7 (slow)
Q4 yes 510k 276.4-329.4 t/s ~5.5-6.6 (slow)

Estimate (memory-bound). Beyond "max context" the KV cache spills to system RAM offload (depends on your system RAM) and speed drops to the "beyond" figure.

FAQ

Can the NVIDIA H100 PCIe 80GB run Generative BST?โ–ถ

Yes. At Q4 it generates ~276.4-329.4 tok/s and fits up to 510k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~5.5-6.6 tok/s.

How much context fits?โ–ถ

At Q4, up to about 510k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.