GPUs / Apple

Apple M3 Max

Memory
LPDDR5
Bus
512-bit
Published peak
400 GB/s
Capacity
from 36 GB

Measured

Read
275.5GB/s
Write
220.0GB/s
Copy
240.1GB/s
Random
52.5GB/s
Latency
394ns
Sustain
240.9GB/s

Medians of 1 verified run from 1 operator.

Local AI speed

ModelParameters4-bit, tokens/s8-bit, tokens/s
Llama 3.1 8B8.03B≤ 69≤ 34
Qwen 2.5 14B14.7B≤ 37≤ 19
gpt-oss-20b21B, 3.6B active≤ 153≤ 77
Mistral Small 24B24B≤ 23≤ 11
Qwen 2.5 32B32.5B≤ 17≤ 8.5
Llama 3.3 70B70.6B≤ 7.8Does not fit
gpt-oss-120b117B, 5.1B activeDoes not fitDoes not fit
Upper bounds: median read ÷ bytes read per token. Mixture-of-experts models read only their active experts. Fit checks weights against the smallest 36 GB configuration; the KV cache needs room on top.

Top operators

#OperatorScoreReadDate
1OP-FF606A69.6275.5 GB/s2026-09-27

Full Apple M3 Max leaderboard