MulaiPapan peringkatDecode calculatorModelOrganizationsReportsMatrix runsHardwareTolok UkurMarketplaceSewaProDokumen API
Bahasa
Actual Computer — Every computer, one endpoint

Speed test matrix

Qwen3.5-9B on H200 NVL 141GB

sglang · 0.5.6.post3.dev9160+g1cf2b8c54 · BF166 contexts (8K–251K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K, 251K, 256K
by @keennaylmx dev importMeasured

Hardware

Type
Discrete GPU
GPU
H200 NVL
VRAM
141 GB
OS
linux

Engine

Engine
sglang
Version
0.5.6.post3.dev9160+g1cf2b8c54
Quantization
BF16
Backend
cuda
Build
cuda-13.2
Commit
1cf2b8c54d81

Run

Served model
qwen
Revision
main
Decode test
Fixed duration
Prefill mode
Measured during decode warm-up
Time per measurement
30 s
Max output
4,096 tokens

Launch command

"$HOME/env_h200-qwen-sglang/bin/sglang" serve --model-path Qwen/Qwen3.5-9B --served-model-name qwen --trust-remote-code --tp 1 --reasoning-parser qwen3 --tool-call-parser qwen3_coder --context-length 262144 --mem-fraction-static 0.862486 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted> --enable-flashinfer-allreduce-fusion

Prefill

prompt tok/s by context length
8K context27.8K293 ms TTFT
16K context28.9K558 ms TTFT
32K context28.4K1,130 ms TTFT
64K context26.2K2,444 ms TTFT
128K context22.4K5,724 ms TTFT
251K context17K14,718 ms TTFT
256K context17K15,019 ms TTFT

Sustained decode

total tok/s · rows are context length, columns are simultaneous requests
Context1248163264128
8K172.782 ms TTFT263.9100 ms TTFT637.9118 ms TTFT1,152.3192 ms TTFT1,916.8345 ms TTFT2,751.3640 ms TTFT4,434.91,282 ms TTFT4,455.02,443 ms TTFT
16K170.998 ms TTFT87.4153 ms TTFT196.6210 ms TTFT1,048.1313 ms TTFT1,554.5643 ms TTFT2,915.61,279 ms TTFT3,463.32,436 ms TTFTOver capacity
32K167.0161 ms TTFT102.1252 ms TTFT546.1420 ms TTFT755.9788 ms TTFT1,591.91,592 ms TTFT2,113.23,000 ms TTFTOver capacityOver capacity
64K158.9284 ms TTFT98.4524 ms TTFT489.8729 ms TTFT546.41,910 ms TTFT832.03,867 ms TTFTOver capacityOver capacityOver capacity
128K145.7503 ms TTFT100.11,114 ms TTFT358.51,974 ms TTFT516.63,756 ms TTFTOver capacityOver capacityOver capacityOver capacity
251K124.6961 ms TTFT83.91,923 ms TTFT223.43,818 ms TTFTOver capacityOver capacityOver capacityOver capacityOver capacity

Over capacity: not run because this many requests at this context would exceed the server's KV cache.

TTFT is measured after the benchmark warms the server's prompt cache, so it mostly excludes prefill.

All measurements

Latencies are medians. Total tok/s is output across all simultaneous requests.
PhaseContextRequests at onceStatusTotal tok/sPer request tok/sTTFTTime between tokensAverage active requestsCompleted requestsOutput tokensMeasured for
Prefill8K1Valid27.8K—293 ms——1——
Prefill16K1Valid28.9K—558 ms——1——
Prefill32K1Valid28.4K—1,130 ms——1——
Prefill64K1Valid26.2K—2,444 ms——1——
Prefill128K1Valid22.4K—5,724 ms——1——
Prefill251K1Valid17K—14,718 ms——1——
Prefill256K1Valid17K—15,019 ms——1——
Sustained decode8K1Valid172.7173.382 ms5.8 ms1.015,18230.0 s
Sustained decode8K2Valid263.9149.2100 ms6.8 ms2.827,91730.0 s
Sustained decode8K4Valid637.9161.6118 ms6.2 ms5.0419,13430.0 s
Sustained decode8K8Valid1,152.3147.3192 ms6.8 ms9.7834,57030.0 s
Sustained decode8K16Valid1,916.8126.7345 ms7.9 ms21.21657,50330.0 s
Sustained decode8K32Valid2,751.386.7640 ms11.5 ms48.0082,52830.0 s
Sustained decode8K64Valid4,434.967.71,282 ms14.8 ms81.60133,05630.0 s
Sustained decode8K128Valid4,455.035.42,443 ms28.2 ms192.00133,63230.0 s
Sustained decode16K1Valid170.9171.598 ms5.8 ms1.015,12630.0 s
Sustained decode16K2Valid87.442.3153 ms23.7 ms142.802,62030.0 s
Sustained decode16K4Valid196.648.0210 ms20.8 ms121.205,89630.0 s
Sustained decode16K8Valid1,048.1136.1313 ms7.4 ms9.7831,44530.0 s
Sustained decode16K16Valid1,554.598.0643 ms10.2 ms24.0046,62430.0 s
Sustained decode16K32Valid2,915.689.51,279 ms11.2 ms33.2087,45630.0 s
Sustained decode16K64Valid3,463.354.32,436 ms18.4 ms78.90103,90430.0 s
Sustained decode16K128Over capacity————————
Sustained decode32K1Valid167.0167.9161 ms6.0 ms1.015,00930.0 s
Sustained decode32K2Valid102.151.3252 ms19.5 ms66.003,06230.0 s
Sustained decode32K4Valid546.1137.6420 ms7.3 ms8.5416,38230.0 s
Sustained decode32K8Valid755.995.6788 ms10.5 ms12.0022,67230.0 s
Sustained decode32K16Valid1,591.998.31,592 ms10.2 ms16.8047,76030.0 s
Sustained decode32K32Valid2,113.265.93,000 ms15.2 ms34.7063,39230.0 s
Sustained decode32K64Over capacity————————
Sustained decode32K128Over capacity————————
Sustained decode64K1Valid158.9160.6284 ms6.2 ms1.014,76830.0 s
Sustained decode64K2Valid98.449.4524 ms20.3 ms33.702,95230.0 s
Sustained decode64K4Valid489.8131.0729 ms7.7 ms4.8414,69530.0 s
Sustained decode64K8Valid546.470.31,910 ms14.2 ms12.0016,39230.0 s
Sustained decode64K16Valid832.054.03,867 ms18.5 ms24.5024,96030.0 s
Sustained decode64K32Over capacity————————
Sustained decode64K64Over capacity————————
Sustained decode64K128Over capacity————————
Sustained decode128K1Valid145.7148.3503 ms6.7 ms1.014,37230.0 s
Sustained decode128K2Valid100.150.51,114 ms19.8 ms18.003,00230.0 s
Sustained decode128K4Valid358.587.81,974 ms11.4 ms5.5010,75630.0 s
Sustained decode128K8Valid516.665.03,756 ms15.4 ms9.5015,49630.0 s
Sustained decode128K16Over capacity————————
Sustained decode128K32Over capacity————————
Sustained decode128K64Over capacity————————
Sustained decode128K128Over capacity————————
Sustained decode251K1Valid124.6128.8961 ms7.8 ms1.013,73830.0 s
Sustained decode251K2Valid83.942.61,923 ms23.5 ms10.002,51630.0 s
Sustained decode251K4Valid223.458.63,818 ms17.1 ms6.006,70030.0 s
Sustained decode251K8Over capacity————————
Sustained decode251K16Over capacity————————
Sustained decode251K32Over capacity————————
Sustained decode251K64Over capacity————————
Sustained decode251K128Over capacity————————

Over capacity: not run because this many requests at this context would exceed the server's KV cache.