New todaydeepseek-v4.1POST
DeepSeek V4.1 Flash on 4x DGX Spark: SGLang tuned 103 tok/s vs TensorFold 110 tok/s
Decode throughput test: SGLang baseline 71 tps, tuned 103 tps, and TensorFold tuned 110 tps.
dextune
@dextuneArchived 9 hours agoOriginally posted 2026.10.28
dextune @dextuneOriginally posted 2026.10.28
DeepSeek V4.1 Flash 1017 w/ 4x DGX Sparks SGLang vs. TensorFold Decode TPS SGLang Baseline : 71 tps SGLang Tuned : 103 tps TensorFold Tuned : 110 tps - Looks like TensorFold is still cruising along smoothly.
This post quotes another post: “DeepSeek V4.1 Flash on four DGX Sparks: tuned SGLang hits 103 tok/s single-stream”
x.com/dextune/status/2115265469485916225Archived 2026.10.28 12:03