# LLM-Benchmark Konfiguration # Modell : gemma-4-12B-it # Engine : vLLM # Run-ID : run-20260804-052132-ea8471 # GPU : NVIDIA GeForce RTX 3090 Ti # CPU : AMD Ryzen 5 5600X 6-Core Processor # RAM : 30 GB bench@llm-benchmark:~$ /opt/vllm-gemma/venv/bin/python /opt/vllm-gemma/venv/bin/vllm serve /home/godcore/models/gemma4-12b-awq \ --served-model-name gemma-4-12B-it \ --host 192.168.41.116 \ --port 8000 \ --max-model-len 8192 \ --gpu-memory-utilization 0.92 \ --enforce-eager \ --limit-mm-per-prompt '{"image":0}'

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.	vllm
Modellalias?Der Name, unter dem das Modell ueber die API angesprochen wird. Genau dieser Wert muss im Request-Feld 'model' stehen.	gemma-4-12B-it
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.	8192
Alias?Anzeigename des Modells nach aussen (served model name), unabhaengig vom Dateinamen.	gemma-4-12B-it
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.	8192
GPU-Speicher?Anteil des GPU-Speichers (0 bis 1), den vLLM belegen darf. 0.92 = 92 %. Hoeher = mehr Platz fuer den KV-Cache (mehr/laengere parallele Anfragen), aber groesseres Risiko fuer 'Out of Memory'.	0.92
Enforce-Eager?Schaltet die optimierte Graph-Ausfuehrung (CUDA-/HIP-Graphs) AB und rechnet Schritt fuer Schritt. Startet schneller und spart etwas VRAM, ist im laufenden Betrieb aber meist langsamer als mit Graphs.	aktiv
limit-mm-per-prompt?Obergrenze fuer multimodale Eingaben pro Prompt (z.B. Anzahl Bilder). Nur fuer multimodale Modelle relevant.	{"image":0}

All benchmarks of this model To leaderboard

Model comparison

gemma-4-12B-it 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

NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.104,3 tok/sNVIDIA GeForce RTX 5090 1.058,9 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 1.041,8 tok/sNVIDIA RTX A6000 681,8 tok/sNVIDIA GeForce RTX 5070 Ti 677,5 tok/s★ NVIDIA GeForce RTX 3090 Ti 654,9 tok/s this runAMD Radeon PRO W7800 48GB 202,0 tok/sTesla V100-PCIE-32GB 186,2 tok/sNVIDIA GeForce RTX 2060 89,8 tok/sNVIDIA Tesla P100 PCIe 16GB 45,8 tok/s

CPUby processor

AMD Ryzen 9 9950X 16-Core Processor 1.104,3 tok/sAMD Ryzen 7 5800X3D 8-Core Processor 1.058,9 tok/sAMD Ryzen Threadripper PRO 9965WX 24-Cores 1.041,8 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 681,8 tok/sAMD Ryzen Threadripper PRO 5975WX 32-Cores 677,5 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 654,9 tok/s★ AMD Ryzen 5 5600X 6-Core Processor 372,4 tok/s this runAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics 186,2 tok/sIntel(R) Core(TM) i5-7400 CPU @ 3.00GHz 89,8 tok/sAMD Ryzen 9 7945HX with Radeon Graphics 45,8 tok/s

MBby mainboard

ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 1.104,3 tok/sASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 1.058,9 tok/sASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 1.041,8 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 677,5 tok/sMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 654,9 tok/s★ ASUSTeK COMPUTER INC. PRIME A520M-K 372,4 tok/s this runASUSTeK COMPUTER INC. ROG STRIX X570-F GAMING 186,2 tok/sASRock H110 Pro BTC+ 89,8 tok/sShenzhen Meigao Electronic Equipment Co.,Ltd F1FXM (DeskMini Series) 45,8 tok/s

ENGby engine

llama.cpp 1.104,3 tok/s★ vLLM 681,8 tok/s this run

DRVby driver

unbekannt 1.104,3 tok/s

💰 Economics

Economics of this run

Operating cost, TCO and comparison with the next-best runs of the same model at identical concurrency (1× concurrent). Methodology →

⚙️ ConfigurationAll metrics and charts below follow these settings – based on a 24-month runtime.Save to URL Reset

⚡ Electricity price EUR/kWh

⚙️ System utilization 100 %

🖥️ Acquisition EUR

🔌 Idle 25 W

⚡ TDP 450 W

☁️ External LLM (API) show

Input €/1M Output €/1M Input per output token

Electricity0.30 EUR/kWh

Avg power (incl. idle)450 W estimated (TDP)GPU 450 W full load

Avg cost / hourEUR 0.14

Electricity / 1M tokensEUR 0.45

Token / kWh668.48K

Acquisition (system)EUR 1,270 partial priceGPU EUR 999 · CPU EUR 121 · PSU EUR 150

Electricity (2 years)–

TCO (2 years)EUR 3,635

Output tokens (2 years)5.27B

☁️ 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 (25 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

	gemma-4-12B-itNVIDIA GeForce RTX 3090 Ti	gemma-4-12B-itNVIDIA GeForce RTX 5090	gemma-4-12B-itNVIDIA RTX PRO 6000 Blackwell Workstation Edition	gemma-4-12B-it3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition
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)	–	–	–	–

NVIDIA GeForce RTX 3090 Ti · this run 84 tok/s NVIDIA GeForce RTX 5090 154 tok/s NVIDIA RTX PRO 6000 Blackwell Workstation Edition 152 tok/s 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 144 tok/s

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

Marcel Sommer

@marcelsommer