Created bymario-alka.dePowered bygodcore.denoob2claw.detricoma.de
Contributed byMario AlkaQwen (Alibaba)

Qwen2.5-3B-Instruct

Performance benchmark · measured on 08.10.2026 03:42

Benchmark-IDrun-20261008-015402-28ce6d
Timebench 3 - Kombi (Prefill + Generation)Dense3BRuntime: llama.cppQuantisierung: Q4_K_M
ℹ
Public note: Arduino Q (ventunoq) . ARM RK3588 8-Core (4x Cortex-A78C + 4x Cortex-A55) . 14 GB RAM . llama.cpp CPU (Edge-Geraet)
Generation0,07tok/s
Prefill33,07tok/s
Time to First Token360.018,50ms
Total duration1.200,00s
Concurrency10parallel
Ranking in the field
1422of 1427 systems

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

This run is better than 0 % of all comparable systems.
Generation 0,1 tok/s
-100 % vs Ø 425,3
Prefill 33,1 tok/s
-100 % vs Ø 6.659,3
Time to First Token 360.019 ms
+685 % vs Ø 45.848
Distribution in the field0 – 2.491 tok/s
Ø 425 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

gemma-4-E2B-itNVIDIA GeForce RTX 5090 · run-20260728-184455-e28775
2.491,2 tok/s
gemma-4-E2B-itNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181606-b37ed9
2.414,4 tok/s
Nemotron-3.5-Lightning-30B-A3BNVIDIA GeForce RTX 3090 Ti · run-20261021-093815-dce558
2.263,7 tok/s
Nemotron-3-Nano-4BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181607-4cca42
2.182,7 tok/s
gemma-4-E2B-it3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-184456-d87c73
2.143,6 tok/s
gpt-oss-20bNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260730-035052-9c966e
1.995,4 tok/s
gpt-oss-20bNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260728-140954-8007c4
1.993,5 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-0aed86
1.937,7 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-8db0fa
1.911,6 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260727-181606-e0eafd
1.897,0 tok/s
gpt-oss-20bNVIDIA GeForce RTX 5090 · run-20260729-032121-7c961d
1.890,7 tok/s
Nemotron-3-Nano-4BNVIDIA GeForce RTX 5090 · run-20260728-194135-2ed9dd
1.878,7 tok/s
gpt-oss-20b3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-034052-26063b
1.835,7 tok/s
Nemotron-3-Nano-4B3× 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-a77046
1.830,3 tok/s
Qwen2.5-3B-Instruct this runKeine GPU (CPU-only) · run-20261008-015402-28ce6d
0,1 tok/s

How does this benchmark compare on other GPUs?

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

Hardware

GPU: Keine GPU (CPU-only)
CPU: 4 x Cortex-A78C + 4 x Cortex-A55
RAM: 15 GB

Setup

Runtime: llama.cpp
Quantization: Q4_K_M
Model: Qwen2.5-3B-Instruct

Configuration

benchmark-konfiguration — run-20261008-015402-28ce6d
# LLM-Benchmark Konfiguration # Modell : Qwen2.5-3B-Instruct # Engine : llama.cpp # Run-ID : run-20261008-015402-28ce6d # GPU : Keine GPU (CPU-only) # CPU : 4 x Cortex-A78C + 4 x Cortex-A55 # RAM : 15 GB bench@llm-benchmark:~$ llama-server \ -m qwen2.5-3b-instruct-q4_k_m.gguf \ --alias Qwen2.5-3B-Instruct \ -ngl 0 \ -t 8 \ -c 40960 \ -np 10 '(ARM' RK3588 'CPU)'
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.Qwen2.5-3B-Instruct
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.40960
Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird.qwen2.5-3b-instruct-q4_k_m.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.0
Threads?Anzahl CPU-Threads fuer die Token-Generierung (Decode).8
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.40960
np10
execution_typelocal

All benchmarks of this model To leaderboard

Anzeige
Model comparison

Qwen2.5-3B-Instruct 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

4733552361180,005.88311.76617.649Prefill (tok/s)Generation (tok/s)AMD Radeon AI PRO R9700 - 364,1 tok/s Generation, 11.369 tok/s Prefill, TTFT 1.151 ms (3 Laufe)AMD Radeon AI PRO R97...AMD Radeon RX 7900 XTX - 358,2 tok/s Generation, 14.238 tok/s Prefill, TTFT 1.134 ms (9 Laufe)AMD Radeon RX 7900 XTXNVIDIA GeForce RTX 2060 - 259,9 tok/s Generation, 6.911 tok/s Prefill, TTFT 1.971 ms (6 Laufe)NVIDIA GeForce RTX 20...CPU-only - 3,5 tok/s Generation, 30 tok/s Prefill, TTFT 277.455 ms (3 Laufe) | DIESER LAUF★ CPU-only
AMD Radeon AI PRO R9700 364,1 tok/sAMD Radeon RX 7900 XTX 358,2 tok/sNVIDIA GeForce RTX 2060 259,9 tok/s★ CPU-only 3,5 tok/s this run

CPUby processor

4733552361180,006.47312.94619.419Prefill (tok/s)Generation (tok/s)AMD Ryzen 3 3100 4-Core Processor - 364,1 tok/s Generation, 11.369 tok/s Prefill, TTFT 1.151 ms (3 Laufe)AMD Ryzen 3 3100 4-Co...AMD Ryzen Threadripper PRO 3955WX 16-Cores - 358,2 tok/s Generation, 15.665 tok/s Prefill, TTFT 1.104 ms (6 Laufe)AMD Ryzen Threadrippe...Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz - 333,2 tok/s Generation, 8.401 tok/s Prefill, TTFT 1.712 ms (9 Laufe)Intel(R) Xeon(R) CPU ...4 x Cortex-A78C + 4 x Cortex-A55 - 3,5 tok/s Generation, 30 tok/s Prefill, TTFT 277.455 ms (3 Laufe) | DIESER LAUF★ 4 x Cortex-A78C + 4 x...
AMD Ryzen 3 3100 4-Core Processor 364,1 tok/sAMD Ryzen Threadripper PRO 3955WX 16-Cores 358,2 tok/sIntel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz 333,2 tok/s★ 4 x Cortex-A78C + 4 x Cortex-A55 3,5 tok/s this run

MBby mainboard

4733552361180,006.47312.94619.419Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. ROG STRIX X570-F GAMING - 364,1 tok/s Generation, 11.369 tok/s Prefill, TTFT 1.151 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 358,2 tok/s Generation, 15.665 tok/s Prefill, TTFT 1.104 ms (6 Laufe)ASUSTeK COMPUTER INC....Dell Inc. PowerEdge R820 - 333,2 tok/s Generation, 8.401 tok/s Prefill, TTFT 1.712 ms (9 Laufe)Dell Inc. PowerEdge R...unbekannt - 3,5 tok/s Generation, 30 tok/s Prefill, TTFT 277.455 ms (3 Laufe)unbekannt
ASUSTeK COMPUTER INC. ROG STRIX X570-F GAMING 364,1 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 358,2 tok/sDell Inc. PowerEdge R820 333,2 tok/sunbekannt 3,5 tok/s

ENGby engine

4023813603393185.80210.51615.22919.943Prefill (tok/s)Generation (tok/s)vLLM - 355,9 tok/s Generation, 17.308 tok/s Prefill, TTFT 1.253 ms (3 Laufe)vLLMllama.cpp - 364,1 tok/s Generation, 8.438 tok/s Prefill, TTFT 47.449 ms (18 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 364,1 tok/s this runvLLM 355,9 tok/s

DRVby driver

4013823643463289.1229.5119.89910.287Prefill (tok/s)Generation (tok/s)unbekannt - 364,1 tok/s Generation, 9.705 tok/s Prefill, TTFT 40.850 ms (21 Laufe)unbekannt
unbekannt 364,1 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 %
🔌 Idle 0 W
⚡ TDP 10 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)0 W missing
Avg cost / hour–
Electricity / 1M tokens–
Token / kWh–
Acquisition (system)– partial price
Electricity (2 years)–
TCO (2 years)–
Output tokens (2 years)4.42M
☁️ 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 (0 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

Qwen2.5-3B-InstructKeine GPU (CPU-only)Qwen2.5-3B-InstructAMD Radeon AI PRO R9700Qwen2.5-3B-Instruct2x AMD Radeon RX 7900 XTXQwen2.5-3B-Instruct2x AMD Radeon RX 7900 XTX
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.