BREAKING
Gemma 4 26B: 16 Instances on DGX Spark
0
x
instances
0
tok/s
aggregate
0
x
scalable to
0
B
total params
0
B
active
0
experts
0
GB
unified memory
0
GB/s
bandwidth
0
$
from
Single vs Parallel Speed
single
52
per-instance
18
Local Open Models Go Workstation
AI NEWS BLITZ
Google's Gemma team showed sixteen models running at once on a single DGX Spark.