H100 80GB vs M1 Max 32GB

For running local LLMs · 2 reports across 2 models

Side A

Vendor: nvidia
VRAM: 80GB
Memory: Discrete

Side B

Vendor: apple
VRAM: 32GB
Memory: Unified

Tokens per second by model

Model	H100 80GB	M1 Max 32GB
Gemma 4up to 31B	—	10.0n=1
Qwen3.6up to 35B	45.0n=1	—

More comparisons