llamaperf

A100 40GB vs H100 80GB

For running local LLMs · 6 reports across 5 models

Side A
A100 40GB
Vendor
nvidia
VRAM
40GB
Memory
Discrete
Side B
H100 80GB
Vendor
nvidia
VRAM
80GB
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 →

ModelA100 40GBH100 80GB
DeepSeek V4 Flashup to 285B16.1n=1193.0n=1
Gemma 4up to 31B125.3n=1
Qwen3.5up to 122Bn=1
Qwen3.6up to 35B45.0n=1
Qwen3.8up to 2400Bn=1

More comparisons