llamaperf

DGX Spark vs V100 32GB

For running local LLMs · 26 reports across 9 models

Side A
DGX Spark
Vendor
nvidia
VRAM
128GB
Memory
Unified
Side B
V100 32GB
Vendor
nvidia
VRAM
32GB
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 →

ModelDGX SparkV100 32GB
DeepSeek V4 Flashup to 285B90.0n=10
Qwen3.8up to 2400B181.0n=365.7n=2
Ling-3.0up to 124B40.9n=3
GLM-5.2up to 744B24.0n=2
Qwen3.6up to 35Bn=182.0n=1
DeepSeek V3n=1
GLM-5.1n=1
Kimi K320.0n=1
Museup to 30B37.0n=1

More comparisons