phi-4-4bit

on M3 Pro

15.6

tok/sdecode

M3 Pro (18-core GPU) + 36GB unified

rank in tier

8/11M3 Pro runstop 73%

best workload

chat-short

where the rig flew

slowest workload

—

single-workload run

backend

mlxMLX is the fastest backend on Apple Silicon for dense models.

faster than

Qwen3-32B-Instruct.Q4_K_M on smoke-host10.0 tok/s
m on x10.0 tok/s

This is a shareable view of one signed llm-speed suite-suite-v1 run. Numbers are not edited. The raw record has every workload, the public key, and the fingerprint hash.