// RUN โ€” FITS

Can the Intel Arc Pro B70 run Japanese-LM-3.6B?

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

Japanese-LM-3.6B on the Intel Arc Pro B70

QuantFits?Max contextSpeedBeyond context
FP16 yes 316k 49-60.8 t/s ~3.2-4 (slow)
Q8 yes 371k 98-121.6 t/s ~6.4-8 (slow)
Q4 yes 396k 174.9-217.1 t/s ~11.5-14.3 (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 Intel Arc Pro B70 run Japanese-LM-3.6B?โ–ถ

Yes. At Q4 it generates ~174.9-217.1 tok/s and fits up to 396k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~11.5-14.3 tok/s.

How much context fits?โ–ถ

At Q4, up to about 396k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.