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 Aug 29, 2026, 11:51 UTC

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
Repository: github.com/sgl-project/sglang
Build: cuda-13.2
Commit: 834400705f2d

Run

Served model: poolside
Revision: main
Benchmark tool: llm-inference-bench v0.4.29
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

Context	1	2	4	8	16	32	64	128
8K	176.465 ms TTFT	201.895 ms TTFT	544.9102 ms TTFT	690.6128 ms TTFT	1,286.0193 ms TTFT	1,861.9355 ms TTFT	Over capacity	Over capacity
16K	173.691 ms TTFT	106.8105 ms TTFT	451.1144 ms TTFT	821.0191 ms TTFT	1,270.9379 ms TTFT	Over capacity	Over capacity	Over capacity
32K	168.4122 ms TTFT	203.8164 ms TTFT	501.4241 ms TTFT	576.1410 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	160.0165 ms TTFT	194.2263 ms TTFT	440.6392 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	144.9265 ms TTFT	244.2417 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
251K	122.2483 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over 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.

Phase	Context	Requests at once	Status	Total tok/s	Per request tok/s	TTFT	Time between tokens	Average active requests	Completed requests	Output tokens	Measured for
Prefill	8K	1	Valid	34.8K	—	236 ms	—	—	1	—	—
Prefill	16K	1	Valid	34.7K	—	467 ms	—	—	1	—	—
Prefill	32K	1	Valid	32.5K	—	991 ms	—	—	1	—	—
Prefill	64K	1	Valid	26.4K	—	2,436 ms	—	—	1	—	—
Prefill	128K	1	Valid	19.1K	—	6,723 ms	—	—	1	—	—
Prefill	251K	1	Valid	12.4K	—	20,307 ms	—	—	1	—	—
Prefill	256K	1	Valid	12.3K	—	20,892 ms	—	—	1	—	—
Sustained decode	8K	1	Valid	176.4	176.9	65 ms	5.7 ms	1.0	1	5,293	30.0 s
Sustained decode	8K	2	Valid	201.8	101.0	95 ms	9.9 ms	3.0	0	6,054	30.0 s
Sustained decode	8K	4	Valid	544.9	136.2	102 ms	7.3 ms	4.0	4	16,344	30.0 s
Sustained decode	8K	8	Valid	690.6	86.6	128 ms	11.5 ms	12.0	0	20,720	30.0 s
Sustained decode	8K	16	Valid	1,286.0	79.3	193 ms	12.6 ms	18.1	0	38,576	30.0 s
Sustained decode	8K	32	Valid	1,861.9	58.0	355 ms	17.2 ms	41.1	0	55,840	30.0 s
Sustained decode	8K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	8K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	173.6	174.2	91 ms	5.7 ms	1.0	1	5,206	30.0 s
Sustained decode	16K	2	Valid	106.8	53.6	105 ms	18.7 ms	34.0	0	3,202	30.0 s
Sustained decode	16K	4	Valid	451.1	110.9	144 ms	9.0 ms	5.1	0	13,532	30.0 s
Sustained decode	16K	8	Valid	821.0	102.9	191 ms	9.7 ms	8.0	0	24,624	30.0 s
Sustained decode	16K	16	Valid	1,270.9	77.7	379 ms	12.9 ms	16.3	0	38,128	30.0 s
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	168.4	169.3	122 ms	5.9 ms	1.0	1	5,052	30.0 s
Sustained decode	32K	2	Valid	203.8	96.7	164 ms	10.3 ms	10.5	0	6,114	30.0 s
Sustained decode	32K	4	Valid	501.4	125.3	241 ms	8.0 ms	4.0	4	15,039	30.0 s
Sustained decode	32K	8	Valid	576.1	72.4	410 ms	13.8 ms	12.0	0	17,280	30.0 s
Sustained decode	32K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	1	Valid	160.0	161.0	165 ms	6.2 ms	1.0	1	4,799	30.0 s
Sustained decode	64K	2	Valid	194.2	94.2	263 ms	10.6 ms	8.4	0	5,824	30.0 s
Sustained decode	64K	4	Valid	440.6	106.1	392 ms	9.4 ms	4.3	0	13,216	30.0 s
Sustained decode	64K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	1	Valid	144.9	146.4	265 ms	6.8 ms	1.0	1	4,345	30.0 s
Sustained decode	128K	2	Valid	244.2	119.5	417 ms	8.4 ms	2.0	0	7,326	30.0 s
Sustained decode	128K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	251K	1	Valid	122.2	125.1	483 ms	8.0 ms	1.0	1	3,667	30.0 s
Sustained decode	251K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	251K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	251K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	251K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	251K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	251K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	251K	128	Over capacity	—	—	—	—	—	—	—	—

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