gpu4.pics
data via LocalScore
NVIDIA H100
80 GB VRAM NVIDIA 700 W TDP
running Llama 3.2 1B Instruct· Q4_K_M· 1.5B
335
tokens / sec
4,512 · Excellent #5 of 768
Generation
335tok/s
higher is better
Prompt / prefill
18,590tok/s
higher is better
Time to first token
68 ms
lower is better
Composite score
4,512
Model size
1.5B
Runtime
gpu4.pics/r/ls-nvidia-h100-llama-32-1b-instruct-q4-k-m

All benchmarks on this rig

ModelQuantGenPromptTTFTPowerPerf/W$/MtokScore
Llama 3.2 1B Instruct1.5BQ4_K_M33518,59068 ms4,512

How it stacks up — Llama 3.2 1B Instruct

NVIDIA H100
335
NVIDIA L40S
355
NVIDIA RTX 3090 Ti
354
NVIDIA RTX 3090
330
NVIDIA RTX A6000
315
NVIDIA A100 80GB
308