llamaperf

M4 Pro 48GB vs M5 Max 128GB

For running local LLMs · 23 reports across 11 models

Side A
M4 Pro 48GB
Vendor
apple
VRAM
48GB
Memory
Unified
Side B
M5 Max 128GB
Vendor
apple
VRAM
128GB
Memory
Unified

Highest reported generation speed by model family

These maxima come from different community setups. Model size, quantization, GPU count, offloading and concurrency may differ. They do not establish which GPU is faster for the same workload. Check the comparison conditions →

ModelM4 Pro 48GBM5 Max 128GB
Qwen3.8up to 2400B20.3n=155.0n=5
Qwen3.6up to 35B79.4n=4
Gemma 4up to 31B7.5n=3
GLM-5.3up to 321B1.8n=124.0n=1
Museup to 30B18.0n=2
DeepSeek V4 Flashup to 285Bn=1
LingBot-World-V2up to 1Bn=1
Nex-N2.5-mini133.6n=1
Qwen3.5up to 122Bn=1
StepFun 3.733.9n=1
Tencent-HY3up to 295B32.4n=1

More comparisons