// RUN โ€” FITS

Can the NVIDIA H200 run UniRep?

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

UniRep on the NVIDIA H200

QuantFits?Max contextSpeedBeyond context
FP16 yes 1956k 96263.7-114725.3 t/s ~802.2-956 (slow)
Q8 yes 1956k 192527.5-229450.5 t/s ~1604.4-1912.1 (slow)
Q4 yes 1956k 343799.1-409733.1 t/s ~2865-3414.4 (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 H200 run UniRep?โ–ถ

Yes. At Q4 it generates ~343799.1-409733.1 tok/s and fits up to 1956k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~2865-3414.4 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1956k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.