// RUN โ€” FITS

Can the NVIDIA H100 SXM5 80GB run Adaptive Inputs + LayerDrop?

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

Adaptive Inputs + LayerDrop on the NVIDIA H100 SXM5 80GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 1087k 2890.7-3445 t/s ~34.5-41.1 (slow)
Q8 yes 1094k 5781.3-6890.1 t/s ~69-82.3 (slow)
Q4 yes 1097k 10323.8-12303.7 t/s ~123.3-146.9 (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 SXM5 80GB run Adaptive Inputs + LayerDrop?โ–ถ

Yes. At Q4 it generates ~10323.8-12303.7 tok/s and fits up to 1097k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~123.3-146.9 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1097k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.