शुरुआत करेंलीडरबोर्डDecode calculatorमॉडलOrganizationsReportsMatrix runsहार्डवेयरबेंचमार्कमार्केटप्लेसरेंटलProAPI दस्तावेज़
भाषा
Actual Computer — Every computer, one endpoint

Speed test matrix

Laguna-XS.2 on H200 NVL 141GB

sglang · 0.5.6.post3.dev9269+g834400705 · 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.dev9269+g834400705
Quantization
BF16
Backend
cuda
Build
cuda-13.2
Commit
834400705f2d

Run

Served model
poolside
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

env SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1 "$HOME/env_h200-poolside-sglang/bin/sglang" serve --model-path poolside/Laguna-XS.2 --served-model-name poolside --trust-remote-code --tp 1 --reasoning-parser poolside_v1 --tool-call-parser poolside_v1 --context-length 262144 --mem-fraction-static 0.851907 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted>

Prefill

prompt tok/s by context length
8K context34.8K236 ms TTFT
16K context34.7K467 ms TTFT
32K context32.5K991 ms TTFT
64K context26.4K2,436 ms TTFT
128K context19.1K6,723 ms TTFT
251K context12.4K20,307 ms TTFT
256K context12.3K20,892 ms TTFT

Sustained decode

total tok/s · rows are context length, columns are simultaneous requests
Context1248163264128
8K176.465 ms TTFT201.895 ms TTFT544.9102 ms TTFT690.6128 ms TTFT1,286.0193 ms TTFT1,861.9355 ms TTFTOver capacityOver capacity
16K173.691 ms TTFT106.8105 ms TTFT451.1144 ms TTFT821.0191 ms TTFT1,270.9379 ms TTFTOver capacityOver capacityOver capacity
32K168.4122 ms TTFT203.8164 ms TTFT501.4241 ms TTFT576.1410 ms TTFTOver capacityOver capacityOver capacityOver capacity
64K160.0165 ms TTFT194.2263 ms TTFT440.6392 ms TTFTOver capacityOver capacityOver capacityOver capacityOver capacity
128K144.9265 ms TTFT244.2417 ms TTFTOver capacityOver capacityOver capacityOver capacityOver capacityOver capacity
251K122.2483 ms TTFTOver capacityOver capacityOver 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
Prefill8K1Valid34.8K—236 ms——1——
Prefill16K1Valid34.7K—467 ms——1——
Prefill32K1Valid32.5K—991 ms——1——
Prefill64K1Valid26.4K—2,436 ms——1——
Prefill128K1Valid19.1K—6,723 ms——1——
Prefill251K1Valid12.4K—20,307 ms——1——
Prefill256K1Valid12.3K—20,892 ms——1——
Sustained decode8K1Valid176.4176.965 ms5.7 ms1.015,29330.0 s
Sustained decode8K2Valid201.8101.095 ms9.9 ms3.006,05430.0 s
Sustained decode8K4Valid544.9136.2102 ms7.3 ms4.0416,34430.0 s
Sustained decode8K8Valid690.686.6128 ms11.5 ms12.0020,72030.0 s
Sustained decode8K16Valid1,286.079.3193 ms12.6 ms18.1038,57630.0 s
Sustained decode8K32Valid1,861.958.0355 ms17.2 ms41.1055,84030.0 s
Sustained decode8K64Over capacity————————
Sustained decode8K128Over capacity————————
Sustained decode16K1Valid173.6174.291 ms5.7 ms1.015,20630.0 s
Sustained decode16K2Valid106.853.6105 ms18.7 ms34.003,20230.0 s
Sustained decode16K4Valid451.1110.9144 ms9.0 ms5.1013,53230.0 s
Sustained decode16K8Valid821.0102.9191 ms9.7 ms8.0024,62430.0 s
Sustained decode16K16Valid1,270.977.7379 ms12.9 ms16.3038,12830.0 s
Sustained decode16K32Over capacity————————
Sustained decode16K64Over capacity————————
Sustained decode16K128Over capacity————————
Sustained decode32K1Valid168.4169.3122 ms5.9 ms1.015,05230.0 s
Sustained decode32K2Valid203.896.7164 ms10.3 ms10.506,11430.0 s
Sustained decode32K4Valid501.4125.3241 ms8.0 ms4.0415,03930.0 s
Sustained decode32K8Valid576.172.4410 ms13.8 ms12.0017,28030.0 s
Sustained decode32K16Over capacity————————
Sustained decode32K32Over capacity————————
Sustained decode32K64Over capacity————————
Sustained decode32K128Over capacity————————
Sustained decode64K1Valid160.0161.0165 ms6.2 ms1.014,79930.0 s
Sustained decode64K2Valid194.294.2263 ms10.6 ms8.405,82430.0 s
Sustained decode64K4Valid440.6106.1392 ms9.4 ms4.3013,21630.0 s
Sustained decode64K8Over capacity————————
Sustained decode64K16Over capacity————————
Sustained decode64K32Over capacity————————
Sustained decode64K64Over capacity————————
Sustained decode64K128Over capacity————————
Sustained decode128K1Valid144.9146.4265 ms6.8 ms1.014,34530.0 s
Sustained decode128K2Valid244.2119.5417 ms8.4 ms2.007,32630.0 s
Sustained decode128K4Over capacity————————
Sustained decode128K8Over capacity————————
Sustained decode128K16Over capacity————————
Sustained decode128K32Over capacity————————
Sustained decode128K64Over capacity————————
Sustained decode128K128Over capacity————————
Sustained decode251K1Valid122.2125.1483 ms8.0 ms1.013,66730.0 s
Sustained decode251K2Over capacity————————
Sustained decode251K4Over capacity————————
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.