// RUN โ€” FITS

Can the Apple M3 Max 36GB run MegaMolBART?

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

MegaMolBART on the Apple M3 Max 36GB

QuantFits?Max contextSpeedBeyond context
FP16 yes 470k 2266.7-2733.3 t/s ~75.6-91.1 (slow)
Q8 yes 471k 4533.3-5466.7 t/s ~151.1-182.2 (slow)
Q4 yes 471k 8095.2-9761.9 t/s ~269.8-325.4 (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 M3 Max 36GB run MegaMolBART?โ–ถ

Yes. At Q4 it generates ~8095.2-9761.9 tok/s and fits up to 471k context; beyond that it spills to SSD swap and slows to ~269.8-325.4 tok/s.

How much context fits?โ–ถ

At Q4, up to about 471k tokens stay in fast memory; longer context spills to SSD swap and throughput collapses.