// RUN โ€” FITS

Can the NVIDIA B200 run Flan T5-XXL + BLIP-2?

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

Flan T5-XXL + BLIP-2 on the NVIDIA B200

QuantFits?Max contextSpeedBeyond context
FP16 yes 921k 241.3-287.6 t/s ~1.2-1.4 (slow)
Q8 yes 995k 482.6-575.2 t/s ~2.4-2.9 (slow)
Q4 yes 1028k 861.9-1027.2 t/s ~4.3-5.1 (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 B200 run Flan T5-XXL + BLIP-2?โ–ถ

Yes. At Q4 it generates ~861.9-1027.2 tok/s and fits up to 1028k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~4.3-5.1 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1028k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.