Speed test matrix

Mistral-Medium-3.5-128B on H200 NVL 141GB ×2

sglang · 0.5.6.post3.dev9335+gdd15fb57b · FP86 contexts (8K–251K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K, 251K, 256K

by @keennaylmx dev importMeasured Sep 19, 2026, 16:59 UTC

Hardware

Type: Discrete GPU
GPU: H200 NVL
GPU count: 2
VRAM: 282 GB (2 × 141 GB)
OS: linux

Engine

Engine: sglang
Version: 0.5.6.post3.dev9335+gdd15fb57b
Quantization: FP8
Backend: cuda
Repository: github.com/sgl-project/sglang
Build: cuda-13.2
Commit: dd15fb57b5ef

Run

Served model: mistral
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-mistralai-sglang/bin/sglang" serve --model-path mistralai/Mistral-Medium-3.5-128B --served-model-name mistral --trust-remote-code --tp 2 --reasoning-parser mistral --tool-call-parser mistral --context-length 262144 --mem-fraction-static 0.886040 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted> --dtype bfloat16 --speculative-algorithm EAGLE --speculative-draft-model-path mistralai/Mistral-Medium-3.5-128B-EAGLE --speculative-num-steps 3 --speculative-eagle-topk 1 --speculative-num-draft-tokens 4

Prefill

prompt tok/s by context length

8K context5,2181,576 ms TTFT

16K context4,8513,354 ms TTFT

32K context4,2347,644 ms TTFT

64K context3,27819,686 ms TTFT

128K context2,25057,274 ms TTFT

251K context1,410179,076 ms TTFT

256K context1,391185,130 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	65.9Requests queued69 ms TTFT	66.1146 ms TTFT	95.8185 ms TTFT	185.1239 ms TTFT	627.81,063 ms TTFT	Over capacity	Over capacity	Over capacity
16K	65.9101 ms TTFT	72.7181 ms TTFT	145.4213 ms TTFT	354.5242 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	60.0122 ms TTFT	87.5225 ms TTFT	171.9269 ms TTFT	269.7411 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	59.0169 ms TTFT	67.2265 ms TTFT	107.3405 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	54.4261 ms TTFT	50.8467 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
251K	39.4Requests queued445 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity

Requests queued: the server could not run all requests at once, so some waited in a queue. 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	5,218.0	—	1,576 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,851.0	—	3,354 ms	—	—	1	—	—
Prefill	32K	1	Valid	4,234.0	—	7,644 ms	—	—	1	—	—
Prefill	64K	1	Valid	3,278.0	—	19,686 ms	—	—	1	—	—
Prefill	128K	1	Valid	2,250.0	—	57,274 ms	—	—	1	—	—
Prefill	251K	1	Valid	1,410.0	—	179,076 ms	—	—	1	—	—
Prefill	256K	1	Valid	1,391.0	—	185,130 ms	—	—	1	—	—
Sustained decode	8K	1	Requests queued	65.9	66.0	69 ms	15.2 ms	1.0	0	1,977	30.0 s
Sustained decode	8K	2	Valid	66.1	33.2	146 ms	30.1 ms	3.0	0	1,981	30.0 s
Sustained decode	8K	4	Valid	95.8	23.4	185 ms	42.8 ms	7.0	0	2,872	30.0 s
Sustained decode	8K	8	Valid	185.1	22.7	239 ms	44.0 ms	15.0	0	5,547	30.0 s
Sustained decode	8K	16	Valid	627.8	39.2	1,063 ms	25.5 ms	29.6	0	18,837	30.0 s
Sustained decode	8K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	8K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	8K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	65.9	65.5	101 ms	15.3 ms	1.0	0	1,976	30.0 s
Sustained decode	16K	2	Valid	72.7	36.2	181 ms	27.7 ms	28.6	0	2,179	30.0 s
Sustained decode	16K	4	Valid	145.4	35.6	213 ms	28.1 ms	23.8	0	4,359	30.0 s
Sustained decode	16K	8	Valid	354.5	43.9	242 ms	22.8 ms	13.8	0	10,635	30.0 s
Sustained decode	16K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	60.0	59.7	122 ms	16.8 ms	1.0	0	1,798	30.0 s
Sustained decode	32K	2	Valid	87.5	43.3	225 ms	23.1 ms	13.0	0	2,626	30.0 s
Sustained decode	32K	4	Valid	171.9	41.3	269 ms	24.2 ms	10.8	0	5,155	30.0 s
Sustained decode	32K	8	Valid	269.7	33.8	411 ms	29.6 ms	13.4	0	8,091	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	59.0	58.0	169 ms	17.2 ms	1.0	0	1,771	30.0 s
Sustained decode	64K	2	Valid	67.2	33.4	265 ms	30.1 ms	13.8	0	2,015	30.0 s
Sustained decode	64K	4	Valid	107.3	26.6	405 ms	37.5 ms	14.0	0	3,217	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	54.4	54.0	261 ms	18.5 ms	1.0	0	1,631	30.0 s
Sustained decode	128K	2	Valid	50.8	25.1	467 ms	40.0 ms	12.5	0	1,523	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	Requests queued	39.4	39.3	445 ms	25.4 ms	1.0	0	1,180	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	—	—	—	—	—	—	—	—

Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.