Pierwsze krokiRankingDecode calculatorModeleOrganizationsReportsMatrix runsSprzętBenchmarkiMarketplaceWynajemProDokumentacja API
Język
Actual Computer — Every computer, one endpoint

Speed test matrix

North-Mini-Code-1.0 on H200 NVL 141GB

vllm · 0.26.1rc1.dev638+g52be12cfa · BF167 contexts (8K–314,880) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K, 256K, 314,880, 320,000
by @keennaylmx dev importMeasured

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
Build
cuda-13.2
Commit
52be12cfac0c

Run

Served model
north-mini-code
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

"$HOME/env_h200-cohere-vllm/bin/vllm" serve CohereLabs/North-Mini-Code-1.0 --served-model-name north-mini-code --trust-remote-code --tensor-parallel-size 1 --reasoning-parser cohere_command4 --enable-auto-tool-choice --tool-call-parser cohere_command4 --max-model-len 320000 --gpu-memory-utilization 0.885868 --host 0.0.0.0 --port 8000 --api-key <redacted>

Prefill

prompt tok/s by context length
8K context39.4K207 ms TTFT
16K context36K447 ms TTFT
32K context33.4K957 ms TTFT
64K context28.4K2,248 ms TTFT
128K context21.1K6,021 ms TTFT
256K context13.5K18,805 ms TTFT
314,880 context11.7K26,016 ms TTFT
320,000 context11.7K26,594 ms TTFT

Sustained decode

total tok/s · rows are context length, columns are simultaneous requests
Context1248163264128
8K198.253 ms TTFT338.967 ms TTFT563.698 ms TTFT867.3181 ms TTFT1,347.6228 ms TTFT2,054.3354 ms TTFT2,921.7523 ms TTFT4,032.8937 ms TTFT
16K195.680 ms TTFT334.6107 ms TTFT550.2143 ms TTFT854.3213 ms TTFT1,317.3296 ms TTFT1,918.3576 ms TTFT2,703.2771 ms TTFTOver capacity
32K191.3123 ms TTFT327.0118 ms TTFT532.8225 ms TTFT809.2342 ms TTFT1,210.2543 ms TTFT1,741.9694 ms TTFT2,344.01,359 ms TTFTOver capacity
64K183.9192 ms TTFT308.5235 ms TTFT494.5243 ms TTFT736.7582 ms TTFT1,062.01,009 ms TTFT1,438.91,782 ms TTFTOver capacityOver capacity
128K169.9323 ms TTFT279.6383 ms TTFT442.2571 ms TTFT618.2960 ms TTFT851.41,552 ms TTFTOver capacityOver capacityOver capacity
256K147.5644 ms TTFT237.1881 ms TTFT356.81,214 ms TTFT468.82,051 ms TTFTOver capacityOver capacityOver capacityOver capacity
314,880139.8797 ms TTFT222.81,150 ms TTFT334.41,672 ms TTFTOver 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
Prefill8K1Valid39.4K—207 ms——1——
Prefill16K1Valid36K—447 ms——1——
Prefill32K1Valid33.4K—957 ms——1——
Prefill64K1Valid28.4K—2,248 ms——1——
Prefill128K1Valid21.1K—6,021 ms——1——
Prefill256K1Valid13.5K—18,805 ms——1——
Prefill314,8801Valid11.7K—26,016 ms——1——
Prefill320,0001Valid11.7K—26,594 ms——1——
Sustained decode8K1Valid198.2198.853 ms5.0 ms1.015,94630.0 s
Sustained decode8K2Valid338.9170.867 ms5.9 ms2.0210,16730.0 s
Sustained decode8K4Valid563.6142.798 ms7.0 ms4.0416,90730.0 s
Sustained decode8K8Valid867.3109.9181 ms9.1 ms8.0026,01630.0 s
Sustained decode8K16Valid1,347.685.5228 ms11.7 ms16.0040,41730.0 s
Sustained decode8K32Valid2,054.365.5354 ms15.3 ms32.0061,62930.0 s
Sustained decode8K64Valid2,921.746.3523 ms21.6 ms64.0087,61430.0 s
Sustained decode8K128Valid4,032.831.8937 ms31.4 ms128.00120,95830.0 s
Sustained decode16K1Valid195.6196.480 ms5.1 ms1.015,86930.0 s
Sustained decode16K2Valid334.6168.5107 ms5.9 ms2.0210,03830.0 s
Sustained decode16K4Valid550.2140.7143 ms7.1 ms4.0416,50330.0 s
Sustained decode16K8Valid854.3108.0213 ms9.3 ms8.0025,62430.0 s
Sustained decode16K16Valid1,317.383.1296 ms12.0 ms16.0039,52030.0 s
Sustained decode16K32Valid1,918.360.9576 ms16.4 ms32.0057,53630.0 s
Sustained decode16K64Valid2,703.242.7771 ms23.4 ms64.0081,08230.0 s
Sustained decode16K128Over capacity————————
Sustained decode32K1Valid191.3192.4123 ms5.2 ms1.015,73730.0 s
Sustained decode32K2Valid327.0164.6118 ms6.1 ms2.029,81130.0 s
Sustained decode32K4Valid532.8136.3225 ms7.3 ms3.9415,98730.0 s
Sustained decode32K8Valid809.2102.0342 ms9.8 ms8.0024,28030.0 s
Sustained decode32K16Valid1,210.276.3543 ms13.1 ms16.0036,30430.0 s
Sustained decode32K32Valid1,741.955.0694 ms18.2 ms32.0052,25730.0 s
Sustained decode32K64Valid2,344.036.81,359 ms27.2 ms64.0070,33530.0 s
Sustained decode32K128Over capacity————————
Sustained decode64K1Valid183.9185.6192 ms5.4 ms1.015,51730.0 s
Sustained decode64K2Valid308.5156.4235 ms6.4 ms2.029,25530.0 s
Sustained decode64K4Valid494.5125.4243 ms8.0 ms4.0414,83330.0 s
Sustained decode64K8Valid736.792.6582 ms10.8 ms8.0022,11230.0 s
Sustained decode64K16Valid1,062.066.71,009 ms15.0 ms16.0031,85630.0 s
Sustained decode64K32Valid1,438.945.31,782 ms22.1 ms32.0043,17130.0 s
Sustained decode64K64Over capacity————————
Sustained decode64K128Over capacity————————
Sustained decode128K1Valid169.9172.4323 ms5.8 ms1.015,09630.0 s
Sustained decode128K2Valid279.6142.8383 ms7.0 ms2.028,38730.0 s
Sustained decode128K4Valid442.2110.7571 ms9.0 ms4.0013,26430.0 s
Sustained decode128K8Valid618.277.5960 ms12.9 ms8.0018,55230.0 s
Sustained decode128K16Valid851.453.11,552 ms18.8 ms16.0025,55230.0 s
Sustained decode128K32Over capacity————————
Sustained decode128K64Over capacity————————
Sustained decode128K128Over capacity————————
Sustained decode256K1Valid147.5152.5644 ms6.6 ms1.014,42630.0 s
Sustained decode256K2Valid237.1119.2881 ms8.4 ms2.007,11030.0 s
Sustained decode256K4Valid356.889.71,214 ms11.2 ms4.0010,70930.0 s
Sustained decode256K8Valid468.859.12,051 ms16.9 ms8.0014,06330.0 s
Sustained decode256K16Over capacity————————
Sustained decode256K32Over capacity————————
Sustained decode256K64Over capacity————————
Sustained decode256K128Over capacity————————
Sustained decode314,8801Valid139.8145.7797 ms6.9 ms1.014,19230.0 s
Sustained decode314,8802Valid222.8112.21,150 ms8.9 ms2.006,68430.0 s
Sustained decode314,8804Valid334.484.01,672 ms11.9 ms4.0010,03430.0 s
Sustained decode314,8808Over capacity————————
Sustained decode314,88016Over capacity————————
Sustained decode314,88032Over capacity————————
Sustained decode314,88064Over capacity————————
Sustained decode314,880128Over capacity————————

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