Mac Mini

A Mac Mini ?

Linux

Hardware 2026

What matters for Qwen models, and most/all local MoE models (ie. where the performance is limited) is memory bandwidth. - HN

Here’s the top Apple chips by memory bandwidth (and to steal from clickbait: Apple definitely does not want you to think too closely about this):

Chip Memory Bandwidth
M3 Ultra 819 GB/s
M2 Ultra 800 GB/s
M1 Ultra 800 GB/s
M5 Max (40-core GPU) 610 GB/s
M4 Max (16-core CPU / 40-core GPU) 546 GB/s
M4 Max (14-core CPU / 32-core GPU) 410 GB/s
M2 Max 400 GB/s
M3 Max (16-core CPU / 40-core GPU) 400 GB/s
M1 Max 400 GB/s

LLLM

MacStudio

Mini M4/Pro

Notes

  • The M4 Pro adds more performance cores — roughly double the number of performance cores compared with base M4, leading to higher multi-core throughput.
  • The M4 Pro has significantly more GPU cores — up to 20 cores compared to 8–10 on the base M4.
  • M4 Pro supports much greater memory bandwidth (≈273 GB/s vs 120 GB/s) and larger unified memory, important for heavy pro apps and GPU workloads.
  • M4 SSD are upgradable but not standard

internal

Feature M4 (Base) M4 Pro
CPU Cores (Total) 8–10 12–14
Performance Cores 4 8–10
Efficiency Cores 4–6 4
Max CPU Clock (P-cores) ~4.4 GHz ~4.5 GHz
GPU Cores 8–10 16–20
GPU Architecture Apple GPU (HW ray tracing) Apple GPU (HW ray tracing)
Peak GPU Performance ~3.5–4.4 TFLOPS ~8–9+ TFLOPS
Memory Bandwidth ~120 GB/s ~273 GB/s
Max Unified Memory 32 GB 64 GB
Relative CPU Performance Baseline ~1.5–2× faster (multi-core)
Relative GPU Performance Baseline ~2× faster
Target Workloads Everyday use, light creative work Pro apps, video, 3D, large builds

M2

Written on January 10, 2026, Last update on
LLM pc-hardware