llamaperf

M4 Pro 48GB vs RTX 4090

For running local LLMs · 22 reports across 7 models

Side A
M4 Pro 48GB
Vendor
apple
VRAM
48GB
Memory
Unified
Side B
RTX 4090
Vendor
nvidia
VRAM
24GB
Memory
Discrete

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 48GBRTX 4090
Qwen3.8up to 2400B20.3n=1134.0n=7
DeepSeek V4 Flashup to 285B180.0n=6
Gemma 4up to 31B149.6n=2
Museup to 30B18.0n=2
Qwen3.6up to 35B80.0n=2
GLM-5.3up to 321B1.8n=1
LingBot-World-V2up to 1Bn=1

More comparisons