// RUN โ€” FITS

Can the Intel Gaudi 3 run StarCoder 2 15B?

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

StarCoder 2 15B on the Intel Gaudi 3

QuantFits?Max contextSpeedBeyond context
FP16 yes 329k 71.5-88.8 t/s ~0.8-1 (slow)
Q8 yes 386k 143.1-177.6 t/s ~1.5-1.9 (slow)
Q4 yes 411k 255.5-317.1 t/s ~2.8-3.4 (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 Intel Gaudi 3 run StarCoder 2 15B?โ–ถ

Yes. At Q4 it generates ~255.5-317.1 tok/s and fits up to 411k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~2.8-3.4 tok/s.

How much context fits?โ–ถ

At Q4, up to about 411k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.