⭐ Built in the Agent OS — get it inside the AI Profit BoardroomJoin AIPB →

MLX//SPEEDTEST

apple silicon · unified-memory inference benchmark

T+ 00:00 RUN #01 SIMULATED RUNNING

Models

Results gen tok/s

0

tok/s

STANDBY

—

—gen tok/s

—prefill tok/s

—GB/s

Throughput — rolling window 0 tok/s

Platform

CHIPApple M3 Max

CPU / GPU16-core / 40-core

MEMORY BUS400 GB/s

BACKENDMLX 0.16 · Metal 3

THERMAL● nominal

unified memory — 0.4 / 128 GB resident

Benchmark log live

SPACEpause / resume ←→switch model Rre-run test CLICK GAUGEre-run test workload is simulated — numbers illustrative