gpu4.pics
data via LocalScore
NVIDIA L40S
44 GB VRAM NVIDIA
running Llama 3.2 1B Instruct· Q4_K_M· 1.5B
355
tokens / sec
5,252 · Excellent #4 of 768
Generation
355tok/s
higher is better
Prompt / prefill
24,809tok/s
higher is better
Time to first token
61 ms
lower is better
Composite score
5,252
Model size
1.5B
Runtime
gpu4.pics/r/ls-nvidia-l40s-llama-32-1b-instruct-q4-k-m

All benchmarks on this rig

ModelQuantGenPromptTTFTPowerPerf/W$/MtokScore
Llama 3.2 1B Instruct1.5BQ4_K_M35524,80961 ms5,252

How it stacks up — Llama 3.2 1B Instruct

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