Created bymario-alka.dePowered bygodcore.denoob2claw.detricoma.de
Contributed byMario AlkaStepFun

Step-3.5-Flash

Performance benchmark · measured on 28.07.2026 18:01

Benchmark-IDrun-20260728-184455-5dd873
Timebench 3 - Kombi (Prefill + Generation)Runtime: llama.cpp
Generation739,08tok/s
Prefill4.115,95tok/s
Time to First Token16.441,00ms
Total duration92,38s
Concurrency10parallel
Ranking in the field
16of 41 systems

Performance benchmark · Primary metric: Generation-Speed (tok/s) · 10× concurrent

This run is better than 63 % of all comparable systems.
Generation 739,1 tok/s
-8 % vs Ø 807,0
Prefill 4.116,0 tok/s
-45 % vs Ø 7.506,9
Time to First Token 16.441 ms
-4 % vs Ø 17.148
Distribution in the field55 – 2.144 tok/s
Ø 807 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

gemma-4-E2B-it3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-184456-d87c73
2.143,6 tok/s
NVIDIA-Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-0aed86
1.937,7 tok/s
NVIDIA-Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-8db0fa
1.911,6 tok/s
NVIDIA-Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260727-181606-e0eafd
1.897,0 tok/s
gpt-oss-20b3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-034052-26063b
1.835,7 tok/s
NVIDIA-Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-a77046
1.830,3 tok/s
Laguna-XS-2.13× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-140954-cb60a0
1.615,3 tok/s
Laguna-XS-2.13× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-140954-12a4c5
1.611,0 tok/s
Laguna-XS-2.13× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-140954-593df2
1.606,1 tok/s
North-Mini-Code-1.03× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-164528-6eb96d
1.490,7 tok/s
gemma-4-E4B-it3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194135-d086e1
1.423,3 tok/s
gemma-4-E4B-it3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194135-4155e1
1.421,8 tok/s
gpt-oss-20b3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260724-082940-10761f
1.238,3 tok/s
gpt-oss-120b3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-034051-d52467
1.229,7 tok/s
Step-3.5-Flash this run3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-184455-5dd873
739,1 tok/s

How does this benchmark compare on other GPUs?

Same model on different hardware · 10× concurrent · Generation (tok/s)

Hardware

GPU: 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · 96 GB VRAM
CPU: AMD Ryzen Threadripper PRO 9965WX 24-Cores
RAM: 125 GB
Mainboard: ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE

Setup

Runtime: llama.cpp
Quantization: -
Model: Step-3.5-Flash

Configuration

benchmark-konfiguration — run-20260728-184455-5dd873
# LLM-Benchmark Konfiguration # Modell : Step-3.5-Flash # Engine : llama.cpp # Run-ID : run-20260728-184455-5dd873 # GPU : 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition # CPU : AMD Ryzen Threadripper PRO 9965WX 24-Cores # RAM : 125 GB bench@llm-benchmark:~$ /home/godcore/llama.cpp/build/bin/llama-server \ -m /home/godcore/.cache/huggingface/hub/models--stepfun-ai--Step-3.5-Flash-GGUF-Q4_K_S/snapshots/6930492fee9ffd6f13c2e665a94244c60d2d7d83/step3p5_flash_Q4_K_S-00001-of-00012.gguf \ --alias Step-3.5-Flash \ --host 0.0.0.0 \ --port 8000 \ -ngl 999 \ -c 16384 \ -np 4 \ --jinja
Engine?Die Inferenz-Software, die das Modell ausliefert (z.B. vLLM oder llama.cpp). Sie bestimmt Geschwindigkeit, unterstuetzte Modellformate und welche Parameter ueberhaupt verfuegbar sind.llamacpp
Modellalias?Der Name, unter dem das Modell ueber die API angesprochen wird. Genau dieser Wert muss im Request-Feld 'model' stehen.Step-3.5-Flash
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.16384
Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird./home/godcore/.cache/huggingface/hub/models--stepfun-ai--Step-3.5-Flash-GGUF-Q4_K_S/snapshots/6930492fee9ffd6f13c2e665a94244c60d2d7d83/step3p5_flash_Q4_K_S-00001-of-00012.gguf
GPU-Layer?Anzahl der auf die GPU ausgelagerten Modell-Layer. Hoeher = mehr VRAM und schneller; der Rest laeuft auf der CPU. 999 = alles auf GPU.999
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.16384
np4

All benchmarks of this model To leaderboard

Anzeige
Model comparison

Step-3.5-Flash on various hardware

All published performance runs of this model – each bubble a variant: position = prefill (X) × generation (Y), bubble size = number of runs. Closer to the top right = faster. ★ Marked gold = this benchmark.

GPUby graphics card

9547164772390,001.2592.5173.776Prefill (tok/s)Generation (tok/s)NVIDIA RTX PRO 6000 Blackwell Workstation Edition - 45,5 tok/s Generation, 174 tok/s Prefill, TTFT 129.867 ms (3 Laufe)NVIDIA RTX PRO 6000 B...AMD Radeon AI PRO R9700 - 31,6 tok/s Generation, 298 tok/s Prefill, TTFT 54.060 ms (2 Laufe)AMD Radeon AI PRO R97...NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition - 739,1 tok/s Generation, 3.071 tok/s Prefill, TTFT 7.944 ms (3 Laufe) | DIESER LAUF★ NVIDIA RTX PRO 6000 B...
★ NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 739,1 tok/s this runNVIDIA RTX PRO 6000 Blackwell Workstation Edition 45,5 tok/sAMD Radeon AI PRO R9700 31,6 tok/s

CPUby processor

9547164772390,001.2592.5173.776Prefill (tok/s)Generation (tok/s)AMD Ryzen 9 9950X 16-Core Processor - 45,5 tok/s Generation, 174 tok/s Prefill, TTFT 129.867 ms (3 Laufe)AMD Ryzen 9 9950X 16-...AMD Ryzen Threadripper PRO 7955WX 16-Cores - 31,6 tok/s Generation, 298 tok/s Prefill, TTFT 54.060 ms (2 Laufe)AMD Ryzen Threadrippe...AMD Ryzen Threadripper PRO 9965WX 24-Cores - 739,1 tok/s Generation, 3.071 tok/s Prefill, TTFT 7.944 ms (3 Laufe) | DIESER LAUF★ AMD Ryzen Threadrippe...
★ AMD Ryzen Threadripper PRO 9965WX 24-Cores 739,1 tok/s this runAMD Ryzen 9 9950X 16-Core Processor 45,5 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 31,6 tok/s

MBby mainboard

9527144762380,008001.6012.401Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI - 45,5 tok/s Generation, 174 tok/s Prefill, TTFT 129.867 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE - 739,1 tok/s Generation, 1.962 tok/s Prefill, TTFT 26.391 ms (5 Laufe) | DIESER LAUF★ ASUSTeK COMPUTER INC....
★ ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 739,1 tok/s this runASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 45,5 tok/s

ENGby engine

8137767397026651.2141.2661.3171.369Prefill (tok/s)Generation (tok/s)llama.cpp - 739,1 tok/s Generation, 1.291 tok/s Prefill, TTFT 65.194 ms (8 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 739,1 tok/s this run

DRVby driver

8137767397026651.2141.2661.3171.369Prefill (tok/s)Generation (tok/s)unbekannt - 739,1 tok/s Generation, 1.291 tok/s Prefill, TTFT 65.194 ms (8 Laufe)unbekannt
unbekannt 739,1 tok/s
💰 Economics

Economics of this run

Operating cost, TCO and comparison with the next-best runs of the same model at identical concurrency (10× concurrent). Methodology →

⚙️ ConfigurationAll metrics and charts below follow these settings – based on a 24-month runtime.Save to URLReset
⚡ Electricity price EUR/kWh
⚙️ System utilization 100 %
🖥️ Acquisition EUR
🔌 Idle 165 W
⚡ TDP 983 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)983 W estimated (TDP)GPU 900 + CPU 57 + Board 25 W full load
Avg cost / hourEUR 0.29
Electricity / 1M tokensEUR 0.11
Token / kWh2.71M
Acquisition (system)EUR 45,748 full priceGPU EUR 39,000 · CPU EUR 3,499 · Board EUR 1,299 · RAM EUR 1,750 · PSU EUR 200
Electricity (2 years)
TCO (2 years)EUR 50,912
Output tokens (2 years)46.62B
☁️ External LLM (API) – comparison
External LLM cost (2 years)
Savings vs. external (2 years)

All values above and the charts below take the configured system utilization into account: at X% the system generates only X% of the time, the rest it idles (165 W). Cost per hour drops (more idle), cost per token rises.

Cost over 2 years – electricity only

Cost over 2 years – incl. acquisition (TCO)

Speed vs. tokens per euro

Euro per 1M tokens

Comparison vs. API – economics per benchmark

Step-3.5-Flash3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation EditionStep-3.5-FlashNVIDIA RTX PRO 6000 Blackwell Workstation EditionStep-3.5-Flash3x AMD Radeon AI PRO R9700
Electricity cost (24 mo.)
Acquisition cost
Total cost (TCO)
Generated tokens (24 mo.)
Token price via API
Break-even point (days)
Result (savings / extra cost)

Comparison with up to 3 next-best runs of this model at the same concurrency (at least one on different hardware). Power = GPU TDP + CPU (idle + 15 %) + board (estimated), acquisition = full system (GPU + CPU + board + RAM + PSU), prices = stored market prices.

Contributed by

Mario Alka Administrator

@marioalka

Ich bin Unternehmer, Softwareentwickler und KI-Enthusiast. Seit vielen Jahren entwickle ich Unternehmenssoftware und beschäftige mich inzwischen fast täglich mit lokalen LLMs, KI-Agenten und leistungsfähiger KI-Hardware.

Mit LLM-Benchmark.de möchte ich eine Plattform schaffen, auf der Modelle, GPUs und Agenten objektiv und reproduzierbar miteinander verglichen werden.