llamaperf

H100 80GB vs RTX Pro 6000 Blackwell

For running local LLMs · 32 reports across 10 models

Side A
H100 80GB
Vendor
nvidia
VRAM
80GB
Memory
Discrete
Side B
RTX Pro 6000 Blackwell
Vendor
nvidia
VRAM
96GB
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 →

ModelH100 80GBRTX Pro 6000 Blackwell
Qwen3.8up to 2400Bn=1177.0n=13
DeepSeek V4 Flashup to 285B193.0n=145.0n=4
Qwen3.6up to 35B45.0n=1n=4
Museup to 30B57.0n=2
Aurora1.0up to 0Bn=1
Gemma 4up to 31B125.3n=1
Llama 3.2up to 1Bn=1
Mimo 2.5n=1
Qwen3.5up to 122Bn=1
Tencent-HY3up to 295B57.4n=1

More comparisons