// RUN โ€” FITS

Can the Apple M2 Ultra 64GB run XLNet?

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

XLNet on the Apple M2 Ultra 64GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 846k 800-964.7 t/s ~10-12.1 (slow)
Q8 yes 851k 1600-1929.4 t/s ~20-24.1 (slow)
Q4 yes 853k 2857.1-3445.4 t/s ~35.7-43.1 (slow)

Estimate (memory-bound). Beyond "max context" the KV cache spills to SSD swap and speed drops to the "beyond" figure.

FAQ

Can the Apple M2 Ultra 64GB run XLNet?โ–ถ

Yes. At Q4 it generates ~2857.1-3445.4 tok/s and fits up to 853k context; beyond that it spills to SSD swap and slows to ~35.7-43.1 tok/s.

How much context fits?โ–ถ

At Q4, up to about 853k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.