Contributed byMario AlkaQwen (Alibaba) Qwen3-30B-A3B-Thinking-2507
Performance benchmark · measured on 23.07.2026 04:28
Benchmark-ID
run-20260723-045450-14f178Timebench 3 - Kombi (Prefill + Generation)MoE30BRuntime: vLLMQuantisierung: AWQ
Generation873,72tok/s
Prefill33.022,60tok/s
Time to First Token1.834,50ms
Total duration28,39s
Concurrency10parallel
Ranking in the field
30of 67 systems
Performance benchmark · Primary metric: Generation-Speed (tok/s) · 10× concurrent
This run is better than 56 % of all comparable systems.
Generation
873,7 tok/s
+9 % vs Ø 801,3
Prefill
33.022,6 tok/s
+384 % vs Ø 6.827,5
Time to First Token
1.835 ms
-97 % vs Ø 65.561
Distribution in the field1 – 2.491 tok/s
Wie schlägt sich dieser Benchmark mit anderen Modellen?
gemma-4-E2B-itNVIDIA GeForce RTX 5090 · run-20260728-184455-e28775
2.491,2 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
Laguna-XS-2.1NVIDIA GeForce RTX 5090 · run-20260729-032123-a1af22
1.765,0 tok/s
Laguna-XS-2.1NVIDIA GeForce RTX 5090 · run-20260729-032123-1156eb
1.740,1 tok/s
Laguna-XS-2.1NVIDIA GeForce RTX 5090 · run-20260729-032122-887a06
1.738,4 tok/s
Nemotron-3-Nano-30B-A3BNVIDIA GeForce RTX 5090 · run-20260729-105335-23842b
1.722,1 tok/s
Nemotron-Cascade-2-30B-A3BNVIDIA GeForce RTX 5090 · run-20260729-105335-34f830
1.719,2 tok/s
Qwen3-Coder-30B-A3B-InstructNVIDIA GeForce RTX 5090 · run-20260729-105335-622167
1.716,6 tok/s
Nemotron-3-Nano-Omni-30B-A3B-ReasoningNVIDIA GeForce RTX 5090 · run-20260729-105334-08a7c9
1.706,7 tok/s
Qwen3-30B-A3B-Thinking-2507 same modelNVIDIA GeForce RTX 5090 · run-20260729-105335-e6f72a
1.683,8 tok/s
Qwen3-Omni-30B-A3B-ThinkingNVIDIA GeForce RTX 5090 · run-20260729-105336-ee142c
1.673,6 tok/s
Qwen3-30B-A3B-Instruct-2507NVIDIA GeForce RTX 5090 · run-20260729-105335-2c186b
1.671,0 tok/s
Qwen3-VL-30B-A3B-InstructNVIDIA GeForce RTX 5090 · run-20260729-105336-473379
1.655,9 tok/s
Qwen3-30B-A3B-Thinking-2507 this runNVIDIA GeForce RTX 5090 · run-20260723-045450-14f178
873,7 tok/s
How does this benchmark compare on other GPUs?
Same model on different hardware · 10× concurrent · Generation (tok/s)
NVIDIA GeForce RTX 5090run-20260729-105335-e6f72a
1.683,8 tok/s
NVIDIA RTX PRO 6000 Blackwell Workstation Editionrun-20260729-032123-88fa84
1.670,5 tok/s
3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editionrun-20260729-105342-9f1be6
1.521,4 tok/s
3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editionrun-20260729-105342-a89e7f
1.518,4 tok/s
3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editionrun-20260729-105342-bd5e1f
1.465,3 tok/s
NVIDIA GeForce RTX 3090 Tirun-20260729-060803-f260ab
930,6 tok/s
NVIDIA RTX PRO 6000 Blackwell Workstation Editionrun-20260724-004605-f9a3a1
906,0 tok/s
NVIDIA GeForce RTX 3090 Tirun-20260724-004603-ee41df
883,7 tok/s
NVIDIA GeForce RTX 5090 this runrun-20260723-045450-14f178
873,7 tok/s
2× NVIDIA RTX A6000run-20260829-124021-f57a00
751,7 tok/s
2× NVIDIA RTX A6000run-20260827-090902-07d874
673,2 tok/s
2× NVIDIA RTX A6000run-20260829-181042-e10c46
554,2 tok/s
2× NVIDIA RTX A6000run-20260830-033127-765a91
543,1 tok/s
2× NVIDIA RTX A6000run-20260828-094625-53a7ee
541,2 tok/s
Hardware
GPU: NVIDIA GeForce RTX 5090 · 32 GB VRAM
CPU: AMD Ryzen 7 5800X3D 8-Core Processor
RAM: 126 GB
Mainboard: ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING
Setup
Runtime: vLLM
Quantization: AWQ
Model: Qwen3-30B-A3B-Thinking-2507
Configuration
# LLM-Benchmark Konfiguration
# Modell : Qwen3-30B-A3B-Thinking-2507
# Engine : vLLM
# Run-ID : run-20260723-045450-14f178
# GPU : NVIDIA GeForce RTX 5090
# CPU : AMD Ryzen 7 5800X3D 8-Core Processor
# RAM : 126 GB
bench@llm-benchmark:~$ /opt/gcs_venv/bin/python3 /opt/gcs_venv/bin/vllm serve cyankiwi/Qwen3-30B-A3B-Thinking-2507-AWQ-4bit \
--served-model-name Qwen3-30B-A3B-Thinking-2507 \
--dtype auto \
--max-model-len 8192 \
--gpu-memory-utilization 0.92 \
--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. | Qwen3-30B-A3B-Thinking-2507 |
| 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. | Qwen3-30B-A3B-Thinking-2507 |
| 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'. | 0.92 |
| Trust-Remote-Code?Erlaubt vLLM, mit dem Modell mitgelieferten Python-Code auszufuehren. Noetig fuer Architekturen, die vLLM nicht von Haus aus kennt. Nur bei vertrauenswuerdigen Modellen aktivieren. | aktiv |
Anzeige
Model comparison
Qwen3-30B-A3B-Thinking-2507 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 GeForce RTX 5090 1.683,8 tok/s this runNVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.670,5 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 1.521,4 tok/sNVIDIA GeForce RTX 3090 Ti 930,6 tok/sNVIDIA RTX A6000 751,7 tok/sAMD Radeon PRO W7800 48GB 258,3 tok/sAMD Radeon PRO W7900 Dual Slot 254,6 tok/sNVIDIA GeForce RTX 5070 Ti 155,8 tok/sAMD Radeon AI PRO R9700 91,4 tok/sNVIDIA GeForce RTX 2060 8,5 tok/s
CPUby processor
★ AMD Ryzen 7 5800X3D 8-Core Processor 1.683,8 tok/s this runAMD Ryzen 9 9950X 16-Core Processor 1.670,5 tok/sAMD Ryzen Threadripper PRO 9965WX 24-Cores 1.521,4 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 930,6 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 751,7 tok/sAMD Ryzen Threadripper PRO 5975WX 32-Cores 258,3 tok/sIntel(R) Xeon(R) CPU E5-4620 v2 @ 2.60GHz 8,5 tok/s
MBby mainboard
★ ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 1.683,8 tok/s this runASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 1.670,5 tok/sASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 1.521,4 tok/sMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 930,6 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 258,3 tok/sDell Inc. PowerEdge R820 8,5 tok/s
DRVby driver
unbekannt 1.683,8 tok/sAMD 7.0.0-27-generic 91,4 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
72 W
⚡ TDP
622 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)622 W estimated (TDP)GPU 575 + CPU 35 + Board 12 W full load
Avg cost / hourEUR 0.19
Electricity / 1M tokensEUR 0.059
Token / kWh5.06M
Acquisition (system)EUR 4,986 full priceGPU EUR 3,300 · CPU EUR 349 · Board EUR 149 · RAM EUR 1,008 · PSU EUR 180
Electricity (2 years)–
TCO (2 years)EUR 8,253
Output tokens (2 years)55.11B
☁️ 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 (72 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
| Qwen3-30B-A3B-Thinking-2507NVIDIA GeForce RTX 5090 | Qwen3-30B-A3B-Thinking-2507NVIDIA GeForce RTX 5090 | Qwen3-30B-A3B-Thinking-2507NVIDIA RTX PRO 6000 Blackwell Workstation Edition | Qwen3-30B-A3B-Thinking-25073x 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 5090 · this run 874 tok/sNVIDIA GeForce RTX 5090 1,684 tok/sNVIDIA RTX PRO 6000 Blackwell Workstation Edition 1,670 tok/s3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 1,521 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
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.