Speed test matrix

gemma-4-31B-it on H200 NVL 141GB

vllm · 0.26.1rc1.dev638+g52be12cfa · BF166 contexts (8K–251K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K, 251K, 256K

by @keennaylmx dev importMeasured Sep 3, 2026, 06:11 UTC

Hardware

Type: Discrete GPU
GPU: H200 NVL
VRAM: 141 GB
OS: linux

Engine

Engine: vllm
Version: 0.26.1rc1.dev638+g52be12cfa
Quantization: BF16
Backend: cuda
Repository: github.com/vllm-project/vllm
Build: cuda-13.2
Commit: 52be12cfac0c

Run

Served model: gemma
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

"$HOME/env_h200-gemma-vllm/bin/vllm" serve google/gemma-4-31B-it --served-model-name gemma --trust-remote-code --tensor-parallel-size 1 --reasoning-parser gemma4 --enable-auto-tool-choice --tool-call-parser gemma4 --max-model-len 262144 --gpu-memory-utilization 0.850454 --host 0.0.0.0 --port 8000 --api-key <redacted>

Prefill

prompt tok/s by context length

8K context7,8571,044 ms TTFT

16K context7,5462,155 ms TTFT

32K context6,6074,899 ms TTFT

64K context5,23212,341 ms TTFT

128K context3,71134,756 ms TTFT

251K context2,437103,737 ms TTFT

256K context2,399107,463 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	50.365 ms TTFT	98.5106 ms TTFT	189.3138 ms TTFT	353.2205 ms TTFT	617.8324 ms TTFT	Over capacity	Over capacity	Over capacity
16K	49.8114 ms TTFT	96.8144 ms TTFT	183.8209 ms TTFT	335.3288 ms TTFT	566.5381 ms TTFT	Over capacity	Over capacity	Over capacity
32K	48.8143 ms TTFT	93.5189 ms TTFT	173.9336 ms TTFT	304.9469 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	47.2232 ms TTFT	87.7322 ms TTFT	156.9416 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	44.2474 ms TTFT	77.7633 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
251K	39.31,058 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	7,857.0	—	1,044 ms	—	—	1	—	—
Prefill	16K	1	Valid	7,546.0	—	2,155 ms	—	—	1	—	—
Prefill	32K	1	Valid	6,607.0	—	4,899 ms	—	—	1	—	—
Prefill	64K	1	Valid	5,232.0	—	12,341 ms	—	—	1	—	—
Prefill	128K	1	Valid	3,711.0	—	34,756 ms	—	—	1	—	—
Prefill	251K	1	Valid	2,437.0	—	103,737 ms	—	—	1	—	—
Prefill	256K	1	Valid	2,399.0	—	107,463 ms	—	—	1	—	—
Sustained decode	8K	1	Valid	50.3	50.2	65 ms	19.9 ms	1.0	0	1,508	30.0 s
Sustained decode	8K	2	Valid	98.5	49.3	106 ms	20.3 ms	2.0	0	2,953	30.0 s
Sustained decode	8K	4	Valid	189.3	47.4	138 ms	21.1 ms	4.0	0	5,678	30.0 s
Sustained decode	8K	8	Valid	353.2	44.3	205 ms	22.6 ms	8.0	0	10,591	30.0 s
Sustained decode	8K	16	Valid	617.8	38.8	324 ms	25.8 ms	16.0	0	18,527	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	49.8	49.8	114 ms	20.1 ms	1.0	0	1,492	29.9 s
Sustained decode	16K	2	Valid	96.8	48.4	144 ms	20.7 ms	2.0	0	2,904	30.0 s
Sustained decode	16K	4	Valid	183.8	46.0	209 ms	21.7 ms	4.0	0	5,511	30.0 s
Sustained decode	16K	8	Valid	335.3	42.1	288 ms	23.8 ms	8.0	0	10,056	30.0 s
Sustained decode	16K	16	Valid	566.5	35.5	381 ms	28.1 ms	16.0	0	16,987	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	48.8	48.9	143 ms	20.5 ms	1.0	0	1,464	30.0 s
Sustained decode	32K	2	Valid	93.5	46.8	189 ms	21.3 ms	2.0	0	2,805	30.0 s
Sustained decode	32K	4	Valid	173.9	43.6	336 ms	23.0 ms	4.0	0	5,219	30.0 s
Sustained decode	32K	8	Valid	304.9	38.2	469 ms	26.2 ms	8.0	0	9,149	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	47.2	47.3	232 ms	21.2 ms	1.0	0	1,416	30.0 s
Sustained decode	64K	2	Valid	87.7	44.0	322 ms	22.7 ms	2.0	0	2,633	30.0 s
Sustained decode	64K	4	Valid	156.9	39.3	416 ms	25.4 ms	4.0	0	4,706	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	44.2	44.2	474 ms	22.6 ms	1.0	0	1,324	30.0 s
Sustained decode	128K	2	Valid	77.7	39.0	633 ms	25.7 ms	2.0	0	2,327	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	39.3	39.5	1,058 ms	25.3 ms	1.0	0	1,176	29.9 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.