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

Gemma-4-26B-A4B

Performance benchmark · measured on 26.08.2026 14:07

Benchmark-IDrun-20260826-121549-24e565
Timebench 3 - Kombi (Prefill + Generation)MoE26BRuntime: llama.cppQuantisierung: Q8_0
Generation169,09tok/s
Prefill4.536,78tok/s
Time to First Token2.184,00ms
Total duration66,95s
Concurrency5parallel
Ranking in the field
427of 894 systems

Performance benchmark · Primary metric: Generation-Speed (tok/s) · 5× concurrent

This run is better than 52 % of all comparable systems.
Generation 169,1 tok/s
-37 % vs Ø 268,8
Prefill 4.536,8 tok/s
+14 % vs Ø 3.962,5
Time to First Token 2.184 ms
-94 % vs Ø 39.675
Distribution in the field0 – 1.349 tok/s
Ø 269 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

gemma-4-E2B-itNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181606-0adf6d
1.349,3 tok/s
gemma-4-E2B-itNVIDIA GeForce RTX 5090 · run-20260728-184455-5b4937
1.301,9 tok/s
gemma-4-E2B-it3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-184456-4111a9
1.195,7 tok/s
Nemotron-3-Nano-4BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181607-30d864
1.188,7 tok/s
Nemotron-3-Nano-4BNVIDIA GeForce RTX 5090 · run-20260728-194135-5432cd
1.164,8 tok/s
gpt-oss-20bNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260728-140954-9e81f9
1.136,3 tok/s
gpt-oss-20bNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260730-035052-09627f
1.135,0 tok/s
gpt-oss-20bNVIDIA GeForce RTX 5090 · run-20260729-032121-058a31
1.113,5 tok/s
Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-836a91
1.090,6 tok/s
Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260727-181606-8b381e
1.077,4 tok/s
Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-b8a818
1.071,0 tok/s
Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-038e92
1.051,1 tok/s
gpt-oss-20b3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260729-032121-ffb18a
1.015,5 tok/s
Nemotron-3-Nano-30B-A3BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260729-032121-305ad6
1.007,8 tok/s
Gemma-4-26B-A4B this runNVIDIA RTX A6000 · run-20260826-121549-24e565
169,1 tok/s

Hardware

GPU: NVIDIA RTX A6000 · 48 GB VRAM
CPU: AMD Ryzen Threadripper PRO 7955WX 16-Cores
RAM: 184 GB
Mainboard: ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE

Setup

Runtime: llama.cpp
Quantization: Q8_0
Model: Gemma-4-26B-A4B

Configuration

benchmark-konfiguration — run-20260826-121549-24e565
# LLM-Benchmark Konfiguration # Modell : Gemma-4-26B-A4B # Engine : llama.cpp # Run-ID : run-20260826-121549-24e565 # GPU : NVIDIA RTX A6000 # CPU : AMD Ryzen Threadripper PRO 7955WX 16-Cores # RAM : 184 GB bench@llm-benchmark:~$ /home/godcore/llama.cpp/build/bin/llama-server \ -m gemma-4-26B-A4B-it-Q8_0.gguf \ -ngl 999 \ -fa on \ -c 49152 \ -np 12
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
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.49152
Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird.gemma-4-26B-A4B-it-Q8_0.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
faon
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.49152
np12

All benchmarks of this model To leaderboard

Anzeige
Model comparison

Gemma-4-26B-A4B 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

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)NVIDIA GB10 (DGX Spark) - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)NVIDIA GB10 (DGX Spar...NVIDIA RTX A6000 - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe) | DIESER LAUF★ NVIDIA RTX A6000
★ NVIDIA RTX A6000 264,1 tok/s this runNVIDIA GB10 (DGX Spark) 47,1 tok/s

CPUby processor

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)NVIDIA Grace - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)NVIDIA GraceAMD Ryzen Threadripper PRO 7955WX 16-Cores - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe) | DIESER LAUF★ AMD Ryzen Threadrippe...
★ AMD Ryzen Threadripper PRO 7955WX 16-Cores 264,1 tok/s this runNVIDIA Grace 47,1 tok/s

MBby mainboard

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. GX10 - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe) | DIESER LAUF★ ASUSTeK COMPUTER INC....
★ ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 264,1 tok/s this runASUSTeK COMPUTER INC. GX10 47,1 tok/s

ENGby engine

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)vLLM - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)vLLMllama.cpp - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 264,1 tok/s this runvLLM 47,1 tok/s

DRVby driver

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)unbekannt - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe)unbekanntNVIDIA 590.48.01 / CUDA 13.1 - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)NVIDIA 590.48.01 / CU...
unbekannt 264,1 tok/sNVIDIA 590.48.01 / CUDA 13.1 47,1 tok/s
💰 Economics

Economics of this run

Operating cost, TCO and comparison with the next-best runs of the same model at identical concurrency (5× 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 65 W
⚡ TDP 71 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)71 W missingCPU 46 + Board 25 W full load
Avg cost / hourEUR 0.021
Electricity / 1M tokensEUR 0.035
Token / kWh8.57M
Acquisition (system)EUR 8,394 full priceGPU EUR 3,000 · CPU EUR 1,399 · Board EUR 1,299 · RAM EUR 2,576 · PSU EUR 120
Electricity (2 years)
TCO (2 years)EUR 8,767
Output tokens (2 years)10.66B
☁️ External LLM (API) – comparison
External LLM cost (2 years)
Savings vs. external (2 years)
No power draw measured – values estimated from GPU TDP + CPU (idle + 15 %) + board.

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 (65 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-4-26B-A4BNVIDIA RTX A6000Gemma-4-26B-A4BNVIDIA RTX A6000
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.