Skip to content
Let's Infer
Search models
⌘ K
Search
⌘
Ctrl
K
Appearance
Return to top
All metrics
of
All models
on
DGX Spark
Model
Output speed
(Decode)
Throughput
(Aggregate)
First response
(TTFT)
Nemotron 3.5 Lightning
SGLang
99.33 tok/s
97.58 tok/s
16.73s
Ling 3.0 Flash
SGLang
70.16 tok/s
68.74 tok/s
127.27s
DeepSeek V4 Flash
SparkInfer
56.98 tok/s
37.01 tok/s
181.1s
Qwen3.8 Flash Next
SGLang · 2× NVIDIA DGX Spark
40.45 tok/s
39.92 tok/s
23.38s
Qwen3.8 27B
SGLang
27.33 tok/s
27.15 tok/s
44.51s
