H100 80GB vs M5 32GB

For running local LLMs · 4 reports across 3 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	M5 32GB
Gemma 4up to 31B	125.3n=1	—n=1
Qwen3.5up to 27B	—n=1	—
Qwen3.6up to 35B	45.0n=1	—

More comparisons