Speed test matrix

Kimi-K2.6 on RTX PRO 6000 Blackwell Server Edition 96GB ×8

vllm · 0.30.1rc1.dev452+g73a583112 · INT45 contexts (8K–123K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 123K, 128K

by @keennaylmx dev importMeasured Oct 13, 2026, 09:01 UTC

Hardware

Type: Discrete GPU
GPU: RTX PRO 6000 Blackwell Server Edition
GPU count: 8
VRAM: 768 GB (8 × 96 GB)
OS: linux

Engine

Engine: vllm
Version: 0.30.1rc1.dev452+g73a583112
Quantization: INT4
Backend: cuda
Repository: github.com/vllm-project/vllm
Build: cuda-13.2
Commit: 73a5831127a9

Run

Served model: moonshotai/Kimi-K2.6
Revision: main
Benchmark tool: llm-inference-bench v0.7.9
Decode test: Fixed duration
Prefill mode: Measured during decode warm-up
Time per measurement: 30 s
Max output: 4,096 tokens

Launch command

env VLLM_USE_V2_MODEL_RUNNER=1 FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED=1 VLLM_ENGINE_READY_TIMEOUT_S=3600 "$HOME/env_rtxpro6k-moonshotai-vllm-73a5831127a9/bin/vllm" serve moonshotai/Kimi-K2.6 --served-model-name moonshotai/Kimi-K2.6 --trust-remote-code --tensor-parallel-size 8 --reasoning-parser kimi_k2 --enable-auto-tool-choice --tool-call-parser kimi_k2 --max-model-len 131072 --gpu-memory-utilization 0.91 --host 0.0.0.0 --port 8000 --api-key <redacted> --mm-encoder-tp-mode data --kernel-config.enable_flashinfer_autotune=False --attention-backend TRITON_MLA --moe-backend marlin

Prefill

prompt tok/s by context length

8K context5,8011,411 ms TTFT

16K context5,6692,862 ms TTFT

32K context5,3875,996 ms TTFT

64K context4,87613,220 ms TTFT

123K context4,14129,879 ms TTFT

128K context4,13731,127 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	63.080 ms TTFT	110.2363 ms TTFT	183.2193 ms TTFT	285.4231 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
16K	60.6135 ms TTFT	106.3434 ms TTFT	174.0310 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	57.3228 ms TTFT	100.0567 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	51.3409 ms TTFT	89.8810 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
123K	43.0756 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	5,801.0	—	1,411 ms	—	—	1	—	—
Prefill	16K	1	Valid	5,669.0	—	2,862 ms	—	—	1	—	—
Prefill	32K	1	Valid	5,387.0	—	5,996 ms	—	—	1	—	—
Prefill	64K	1	Valid	4,876.0	—	13,220 ms	—	—	1	—	—
Prefill	123K	1	Valid	4,141.0	—	29,879 ms	—	—	1	—	—
Prefill	128K	1	Valid	4,137.0	—	31,127 ms	—	—	1	—	—
Sustained decode	8K	1	Valid	63.0	63.1	80 ms	15.9 ms	1.0	0	1,890	30.0 s
Sustained decode	8K	2	Valid	110.2	54.7	363 ms	18.3 ms	2.0	0	3,304	30.0 s
Sustained decode	8K	4	Valid	183.2	45.9	193 ms	21.8 ms	4.0	0	5,496	30.0 s
Sustained decode	8K	8	Valid	285.4	35.9	231 ms	27.9 ms	8.0	0	8,558	30.0 s
Sustained decode	8K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	8K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	8K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	8K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	60.6	60.6	135 ms	16.5 ms	1.0	0	1,816	30.0 s
Sustained decode	16K	2	Valid	106.3	52.9	434 ms	18.9 ms	2.0	0	3,190	30.0 s
Sustained decode	16K	4	Valid	174.0	43.8	310 ms	22.8 ms	4.0	0	5,216	30.0 s
Sustained decode	16K	8	Over capacity	—	—	—	—	—	—	—	—
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	57.3	57.4	228 ms	17.4 ms	1.0	0	1,720	30.0 s
Sustained decode	32K	2	Valid	100.0	49.7	567 ms	20.1 ms	2.0	0	2,998	30.0 s
Sustained decode	32K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	8	Over capacity	—	—	—	—	—	—	—	—
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	51.3	51.4	409 ms	19.5 ms	1.0	0	1,539	30.0 s
Sustained decode	64K	2	Valid	89.8	44.5	810 ms	22.5 ms	2.0	0	2,691	30.0 s
Sustained decode	64K	4	Over capacity	—	—	—	—	—	—	—	—
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	123K	1	Valid	43.0	43.0	756 ms	23.2 ms	1.0	0	1,291	30.0 s
Sustained decode	123K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	123K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	123K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	123K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	123K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	123K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	123K	128	Over capacity	—	—	—	—	—	—	—	—

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