llamaperf

M2 Max 96GB vs M3 Ultra 512GB

For running local LLMs · 11 reports across 5 models

Side A
M2 Max 96GB
Vendor
apple
VRAM
96GB
Memory
Unified
Side B
M3 Ultra 512GB
Vendor
apple
VRAM
512GB
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 →

ModelM2 Max 96GBM3 Ultra 512GB
DeepSeek V4 Flashup to 285B20.0n=3
Qwen3.6up to 35B28.0n=3
DeepSeek V4.1 Flashup to 552B31.3n=2
Qwen3.8up to 2400B22.0n=2
Qwen3up to 235B43.0n=1

More comparisons