// RUN โ€” FITS

Can the AMD Instinct MI325X run RWKV-4 World (7B)?

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

RWKV-4 World (7B) on the AMD Instinct MI325X

QuantFits?Max contextSpeedBeyond context
FP16 yes 1673k 255.6-312.5 t/s ~1.7-2.1 (slow)
Q8 yes 1729k 511.3-624.9 t/s ~3.4-4.2 (slow)
Q4 yes 1754k 913-1115.9 t/s ~6.1-7.4 (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 AMD Instinct MI325X run RWKV-4 World (7B)?โ–ถ

Yes. At Q4 it generates ~913-1115.9 tok/s and fits up to 1754k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~6.1-7.4 tok/s.

How much context fits?โ–ถ

At Q4, up to about 1754k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.