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

gemma-4-E2B-it

Performance benchmark · measured on 26.07.2026 21:27

Benchmark-IDrun-20260727-032758-bb6a92
Timebench 3 - Kombi (Prefill + Generation)Dense5BRuntime: vLLM
For context: Teile des Modells wurden per --cpu-offload-gb=80 GB in den System-RAM ausgelagert – RAM wird als erweiterter VRAM genutzt, was den Durchsatz senken kann.
Generation1.608,46tok/s
Prefill31.451,56tok/s
Time to First Token1.632,50ms
Total duration16,69s
Concurrency10parallel
Ranking in the field
3of 76 systems

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

This run is better than 97 % of all comparable systems.
Generation 1.608,5 tok/s
+483 % vs Ø 275,8
Prefill 31.451,6 tok/s
+1.189 % vs Ø 2.440,1
Time to First Token 1.633 ms
-98 % vs Ø 101.921
Distribution in the field2 – 1.744 tok/s
Ø 276 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

How does this benchmark compare on other GPUs?

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

Hardware

GPU: NVIDIA GeForce RTX 5070 Ti · 16 GB VRAM
CPU: AMD Ryzen Threadripper PRO 5975WX 32-Cores
RAM: 247 GB
Mainboard: ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI

Setup

Runtime: vLLM
Quantization: -
Model: gemma-4-E2B-it

Configuration

benchmark-konfiguration — run-20260727-032758-bb6a92
# LLM-Benchmark Konfiguration # Modell : gemma-4-E2B-it # Engine : vLLM # Run-ID : run-20260727-032758-bb6a92 # GPU : NVIDIA GeForce RTX 5070 Ti # CPU : AMD Ryzen Threadripper PRO 5975WX 32-Cores # RAM : 247 GB bench@llm-benchmark:~$ /home/godcore/vllm-venv/bin/python3 /home/godcore/vllm-venv/bin/vllm serve google/gemma-4-E2B-it-qat-w4a16-ct \ --served-model-name gemma-4-E2B-it \ --tensor-parallel-size 1 \ --cpu-offload-gb 80 \ --chat-template /home/godcore/tmpl_gemma.jinja \ --dtype auto \ --max-model-len 8192 \ --gpu-memory-utilization 0.85 \ --trust-remote-code \ --host 0.0.0.0 \ --port 8000
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-E2B-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-E2B-it
Tensor-Parallel?Anzahl GPUs, auf die JEDER einzelne Modell-Layer aufgeteilt wird (Tensor-Parallelitaet). Mehr GPUs = mehr VRAM und meist mehr Speed, aber die Anzahl der Attention-Heads muss durch diesen Wert teilbar sein (z.B. 32 Heads -> nur 1, 2, 4, 8 ... moeglich, NICHT 3).1
cpu-offload-gb?Menge an Modellgewichten in GiB, die in den CPU-RAM ausgelagert wird. Ermoeglicht groessere Modelle als der GPU-Speicher fasst, kostet aber Geschwindigkeit.80
chat-template?Jinja-Vorlage, die Chat-Nachrichten in den Prompt-Text des Modells umwandelt. Noetig, wenn das Modell keine eigene mitbringt./home/godcore/tmpl_gemma.jinja
Dtype?Zahlenformat der Modellgewichte bei der Berechnung (z.B. auto, float16, bfloat16). 'auto' waehlt automatisch das vom Modell empfohlene Format.auto
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'.0.85

All benchmarks of this model To leaderboard

Anzeige
Model comparison

gemma-4-E2B-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

3.2362.4271.6188090,007.11714.23521.352Prefill (tok/s)Generation (tok/s)NVIDIA GeForce RTX 5090 - 2.491,2 tok/s Generation, 11.148 tok/s Prefill, TTFT 2.175 ms (3 Laufe)NVIDIA GeForce RTX 50...NVIDIA RTX PRO 6000 Blackwell Workstation Edition - 2.414,4 tok/s Generation, 16.315 tok/s Prefill, TTFT 1.871 ms (3 Laufe)NVIDIA RTX PRO 6000 B...NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition - 2.143,6 tok/s Generation, 11.441 tok/s Prefill, TTFT 2.226 ms (3 Laufe)NVIDIA RTX PRO 6000 B...NVIDIA RTX A6000 - 1.568,0 tok/s Generation, 17.428 tok/s Prefill, TTFT 1.938 ms (27 Laufe)NVIDIA RTX A6000NVIDIA GeForce RTX 3090 Ti - 1.492,8 tok/s Generation, 13.345 tok/s Prefill, TTFT 1.694 ms (8 Laufe)NVIDIA GeForce RTX 30...NVIDIA GB10 (DGX Spark) - 903,3 tok/s Generation, 6.421 tok/s Prefill, TTFT 7.548 ms (4 Laufe)NVIDIA GB10 (DGX Spar...Intel Arc Pro B70 - 826,2 tok/s Generation, 2.623 tok/s Prefill, TTFT 7.541 ms (3 Laufe)Intel Arc Pro B70AMD Radeon PRO W7800 48GB - 702,9 tok/s Generation, 5.006 tok/s Prefill, TTFT 2.435 ms (11 Laufe)AMD Radeon PRO W7800 ...Tesla V100-PCIE-32GB - 361,8 tok/s Generation, 5.868 tok/s Prefill, TTFT 1.823 ms (6 Laufe)Tesla V100-PCIE-32GBNVIDIA GeForce RTX 2060 - 15,3 tok/s Generation, 1.440 tok/s Prefill, TTFT 8.790 ms (6 Laufe)NVIDIA GeForce RTX 20...NVIDIA GeForce RTX 5070 Ti - 1.744,0 tok/s Generation, 15.451 tok/s Prefill, TTFT 2.253 ms (8 Laufe) | DIESER LAUF★ NVIDIA GeForce RTX 50...
NVIDIA GeForce RTX 5090 2.491,2 tok/sNVIDIA RTX PRO 6000 Blackwell Workstation Edition 2.414,4 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 2.143,6 tok/s★ NVIDIA GeForce RTX 5070 Ti 1.744,0 tok/s this runNVIDIA RTX A6000 1.568,0 tok/sNVIDIA GeForce RTX 3090 Ti 1.492,8 tok/sNVIDIA GB10 (DGX Spark) 903,3 tok/sIntel Arc Pro B70 826,2 tok/sAMD Radeon PRO W7800 48GB 702,9 tok/sTesla V100-PCIE-32GB 361,8 tok/sNVIDIA GeForce RTX 2060 15,3 tok/s

CPUby processor

3.2362.4271.6188090,007.11714.23521.352Prefill (tok/s)Generation (tok/s)AMD Ryzen 7 5800X3D 8-Core Processor - 2.491,2 tok/s Generation, 11.148 tok/s Prefill, TTFT 2.175 ms (3 Laufe)AMD Ryzen 7 5800X3D 8...AMD Ryzen 9 9950X 16-Core Processor - 2.414,4 tok/s Generation, 16.315 tok/s Prefill, TTFT 1.871 ms (3 Laufe)AMD Ryzen 9 9950X 16-...AMD Ryzen Threadripper PRO 9965WX 24-Cores - 2.143,6 tok/s Generation, 11.441 tok/s Prefill, TTFT 2.226 ms (3 Laufe)AMD Ryzen Threadrippe...AMD Ryzen Threadripper PRO 7955WX 16-Cores - 1.568,0 tok/s Generation, 17.428 tok/s Prefill, TTFT 1.938 ms (27 Laufe)AMD Ryzen Threadrippe...AMD Ryzen 9 8945HX with Radeon Graphics - 1.492,8 tok/s Generation, 9.978 tok/s Prefill, TTFT 3.182 ms (3 Laufe)AMD Ryzen 9 8945HX wi...NVIDIA Grace - 903,3 tok/s Generation, 6.421 tok/s Prefill, TTFT 7.548 ms (4 Laufe)NVIDIA GraceAMD Ryzen 9 7945HX with Radeon Graphics - 826,2 tok/s Generation, 2.623 tok/s Prefill, TTFT 7.541 ms (3 Laufe)AMD Ryzen 9 7945HX wi...AMD Ryzen 5 5600X 6-Core Processor - 451,1 tok/s Generation, 15.366 tok/s Prefill, TTFT 801 ms (5 Laufe)AMD Ryzen 5 5600X 6-C...AMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics - 361,8 tok/s Generation, 5.868 tok/s Prefill, TTFT 1.823 ms (6 Laufe)AMD Ryzen 3 PRO 3200G...Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz - 15,3 tok/s Generation, 1.440 tok/s Prefill, TTFT 8.790 ms (6 Laufe)Intel(R) Core(TM) i5-...AMD Ryzen Threadripper PRO 5975WX 32-Cores - 1.744,0 tok/s Generation, 9.404 tok/s Prefill, TTFT 2.358 ms (19 Laufe) | DIESER LAUF★ AMD Ryzen Threadrippe...
AMD Ryzen 7 5800X3D 8-Core Processor 2.491,2 tok/sAMD Ryzen 9 9950X 16-Core Processor 2.414,4 tok/sAMD Ryzen Threadripper PRO 9965WX 24-Cores 2.143,6 tok/s★ AMD Ryzen Threadripper PRO 5975WX 32-Cores 1.744,0 tok/s this runAMD Ryzen Threadripper PRO 7955WX 16-Cores 1.568,0 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 1.492,8 tok/sNVIDIA Grace 903,3 tok/sAMD Ryzen 9 7945HX with Radeon Graphics 826,2 tok/sAMD Ryzen 5 5600X 6-Core Processor 451,1 tok/sAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics 361,8 tok/sIntel(R) Core(TM) i5-7400 CPU @ 3.00GHz 15,3 tok/s

MBby mainboard

3.2362.4271.6188090,006.87013.74020.610Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING - 2.491,2 tok/s Generation, 11.148 tok/s Prefill, TTFT 2.175 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI - 2.414,4 tok/s Generation, 16.315 tok/s Prefill, TTFT 1.871 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE - 2.143,6 tok/s Generation, 16.830 tok/s Prefill, TTFT 1.967 ms (30 Laufe)ASUSTeK COMPUTER INC....Meigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) - 1.492,8 tok/s Generation, 9.978 tok/s Prefill, TTFT 3.182 ms (3 Laufe)Meigao Innovation Tec...ASUSTeK COMPUTER INC. GX10 - 903,3 tok/s Generation, 6.421 tok/s Prefill, TTFT 7.548 ms (4 Laufe)ASUSTeK COMPUTER INC....Shenzhen Meigao Electronic Equipment Co.,Ltd DRFXI (MotherBoard Series) - 826,2 tok/s Generation, 2.623 tok/s Prefill, TTFT 7.541 ms (3 Laufe)Shenzhen Meigao Elect...ASUSTeK COMPUTER INC. PRIME A520M-K - 451,1 tok/s Generation, 15.366 tok/s Prefill, TTFT 801 ms (5 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. ROG STRIX X570-F GAMING - 361,8 tok/s Generation, 5.868 tok/s Prefill, TTFT 1.823 ms (6 Laufe)ASUSTeK COMPUTER INC....ASRock H110 Pro BTC+ - 15,3 tok/s Generation, 1.440 tok/s Prefill, TTFT 8.790 ms (6 Laufe)ASRock H110 Pro BTC+ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 1.744,0 tok/s Generation, 9.404 tok/s Prefill, TTFT 2.358 ms (19 Laufe) | DIESER LAUF★ ASUSTeK COMPUTER INC....
ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 2.491,2 tok/sASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 2.414,4 tok/sASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 2.143,6 tok/s★ ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 1.744,0 tok/s this runMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 1.492,8 tok/sASUSTeK COMPUTER INC. GX10 903,3 tok/sShenzhen Meigao Electronic Equipment Co.,Ltd DRFXI (MotherBoard Series) 826,2 tok/sASUSTeK COMPUTER INC. PRIME A520M-K 451,1 tok/sASUSTeK COMPUTER INC. ROG STRIX X570-F GAMING 361,8 tok/sASRock H110 Pro BTC+ 15,3 tok/s

ENGby engine

2.9172.4832.0501.6161.1833.15612.24721.33930.430Prefill (tok/s)Generation (tok/s)llama.cpp - 2.491,2 tok/s Generation, 7.899 tok/s Prefill, TTFT 3.638 ms (65 Laufe)llama.cppvLLM - 1.608,5 tok/s Generation, 25.687 tok/s Prefill, TTFT 558 ms (17 Laufe) | DIESER LAUF★ vLLM
llama.cpp 2.491,2 tok/s★ vLLM 1.608,5 tok/s this run

DRVby driver

3.0732.3661.6599512441625.0029.84214.682Prefill (tok/s)Generation (tok/s)unbekannt - 2.491,2 tok/s Generation, 12.221 tok/s Prefill, TTFT 2.576 ms (75 Laufe)unbekanntNVIDIA 590.48.01 / CUDA 13.1 - 903,3 tok/s Generation, 6.421 tok/s Prefill, TTFT 7.548 ms (4 Laufe)NVIDIA 590.48.01 / CU...Intel 26.18.38308.4 - 826,2 tok/s Generation, 2.623 tok/s Prefill, TTFT 7.541 ms (3 Laufe)Intel 26.18.38308.4
unbekannt 2.491,2 tok/sNVIDIA 590.48.01 / CUDA 13.1 903,3 tok/sIntel 26.18.38308.4 826,2 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 10 W
⚡ TDP 310 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)310 W estimated (TDP)GPU 300 + Board 10 W full load
Avg cost / hourEUR 0.093
Electricity / 1M tokensEUR 0.016
Token / kWh18.68M
Acquisition (system)EUR 5,000 partial priceGPU EUR 994 · CPU EUR 1,880 · RAM EUR 1,976 · PSU EUR 150
Electricity (2 years)
TCO (2 years)EUR 6,629
Output tokens (2 years)101.45B
☁️ 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 (10 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-E2B-itNVIDIA GeForce RTX 5070 Tigemma-4-E2B-itNVIDIA GeForce RTX 5090gemma-4-E2B-itNVIDIA RTX PRO 6000 Blackwell Workstation Editiongemma-4-E2B-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)

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.