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

Qwen3-1.7B

Performance benchmark · measured on 07.10.2026 22:49

Benchmark-IDrun-20261007-210612-302fef
Timebench 3 - Kombi (Prefill + Generation)Dense1.7BRuntime: 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)
Generation2,95tok/s
Prefill40,70tok/s
Time to First Token60.139,00ms
Total duration817,40s
Concurrency1parallel
Ranking in the field
22of 35 systems

Performance benchmark · Primary metric: Generation-Speed (tok/s) · 1× concurrent

This run is better than 38 % of all comparable systems.
Generation 3,0 tok/s
-43 % vs Ø 5,1
Prefill 40,7 tok/s
-22 % vs Ø 52,0
Time to First Token 60.139 ms
-35 % vs Ø 92.362
Distribution in the field1 – 28 tok/s
Ø 5 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

How does this benchmark compare on other GPUs?

Same model on different hardware · 1× 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: Qwen3-1.7B

Configuration

benchmark-konfiguration — run-20261007-210612-302fef
# LLM-Benchmark Konfiguration # Modell : Qwen3-1.7B # Engine : llama.cpp # Run-ID : run-20261007-210612-302fef # GPU : Keine GPU (CPU-only) # CPU : 4 x Cortex-A78C + 4 x Cortex-A55 # RAM : 15 GB bench@llm-benchmark:~$ llama-server \ -m Qwen3-1.7B-Q4_K_M.gguf \ --alias Qwen3-1.7B \ -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.Qwen3-1.7B
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.Qwen3-1.7B-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

Qwen3-1.7B 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

6404803201600,006.79713.59520.392Prefill (tok/s)Generation (tok/s)AMD Radeon AI PRO R9700 - 492,7 tok/s Generation, 14.046 tok/s Prefill, TTFT 873 ms (3 Laufe)AMD Radeon AI PRO R97...AMD Radeon RX 7900 XTX - 445,2 tok/s Generation, 16.451 tok/s Prefill, TTFT 968 ms (9 Laufe)AMD Radeon RX 7900 XTXAMD Oberon (PlayStation 5) - 202,4 tok/s Generation, 1.591 tok/s Prefill, TTFT 4.381 ms (2 Laufe)AMD Oberon (PlayStati...CPU-only - 4,0 tok/s Generation, 43 tok/s Prefill, TTFT 239.322 ms (3 Laufe) | DIESER LAUF★ CPU-only
AMD Radeon AI PRO R9700 492,7 tok/sAMD Radeon RX 7900 XTX 445,2 tok/sAMD Oberon (PlayStation 5) 202,4 tok/s★ CPU-only 4,0 tok/s this run

CPUby processor

6404803201600,007.76015.52123.281Prefill (tok/s)Generation (tok/s)AMD Ryzen 3 3100 4-Core Processor - 492,7 tok/s Generation, 14.046 tok/s Prefill, TTFT 873 ms (3 Laufe)AMD Ryzen 3 3100 4-Co...AMD Ryzen Threadripper PRO 3955WX 16-Cores - 445,2 tok/s Generation, 18.781 tok/s Prefill, TTFT 935 ms (6 Laufe)AMD Ryzen Threadrippe...Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz - 437,4 tok/s Generation, 11.792 tok/s Prefill, TTFT 1.033 ms (3 Laufe)Intel(R) Xeon(R) CPU ...AMD Zen2 8-Core (PlayStation 5) - 202,4 tok/s Generation, 1.591 tok/s Prefill, TTFT 4.381 ms (2 Laufe)AMD Zen2 8-Core (Play...4 x Cortex-A78C + 4 x Cortex-A55 - 4,0 tok/s Generation, 43 tok/s Prefill, TTFT 239.322 ms (3 Laufe) | DIESER LAUF★ 4 x Cortex-A78C + 4 x...
AMD Ryzen 3 3100 4-Core Processor 492,7 tok/sAMD Ryzen Threadripper PRO 3955WX 16-Cores 445,2 tok/sIntel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz 437,4 tok/sAMD Zen2 8-Core (PlayStation 5) 202,4 tok/s★ 4 x Cortex-A78C + 4 x Cortex-A55 4,0 tok/s this run

MBby mainboard

60047434822195,107.72315.44623.170Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. ROG STRIX X570-F GAMING - 492,7 tok/s Generation, 14.046 tok/s Prefill, TTFT 873 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 445,2 tok/s Generation, 18.781 tok/s Prefill, TTFT 935 ms (6 Laufe)ASUSTeK COMPUTER INC....Dell Inc. PowerEdge R820 - 437,4 tok/s Generation, 11.792 tok/s Prefill, TTFT 1.033 ms (3 Laufe)Dell Inc. PowerEdge R...unbekannt - 202,4 tok/s Generation, 662 tok/s Prefill, TTFT 145.346 ms (5 Laufe)unbekannt
ASUSTeK COMPUTER INC. ROG STRIX X570-F GAMING 492,7 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 445,2 tok/sDell Inc. PowerEdge R820 437,4 tok/sunbekannt 202,4 tok/s

ENGby engine

5635024403793174.83812.31619.79527.273Prefill (tok/s)Generation (tok/s)vLLM - 387,7 tok/s Generation, 23.277 tok/s Prefill, TTFT 1.026 ms (3 Laufe)vLLMllama.cpp - 492,7 tok/s Generation, 8.834 tok/s Prefill, TTFT 52.499 ms (14 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 492,7 tok/s this runvLLM 387,7 tok/s

DRVby driver

54251749346844310.70011.15511.61112.066Prefill (tok/s)Generation (tok/s)unbekannt - 492,7 tok/s Generation, 11.383 tok/s Prefill, TTFT 43.415 ms (17 Laufe)unbekannt
unbekannt 492,7 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 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)186.06M
☁️ 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

Qwen3-1.7BKeine GPU (CPU-only)Qwen3-1.7BAMD Oberon (PlayStation 5)Qwen3-1.7BAMD Radeon AI PRO R9700Qwen3-1.7B2x 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.