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

Qwen3-VL-30B-A3B-Instruct

Performance benchmark · measured on 29.07.2026 02:15

Benchmark-IDrun-20260729-032130-26dd34
Timebench 3 - Kombi (Prefill + Generation)MoE30BRuntime: llama.cppQuantisierung: Q4_K_M
Generation116,57tok/s
Prefill1.336,46tok/s
Time to First Token23.790,50ms
Total duration205,95s
Concurrency5parallel
Ranking in the field
213of 331 systems

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

This run is better than 36 % of all comparable systems.
Generation 116,6 tok/s
-68 % vs Ø 361,9
Prefill 1.336,5 tok/s
-75 % vs Ø 5.274,4
Time to First Token 23.791 ms
-2 % vs Ø 24.397
Distribution in the field0 – 1.349 tok/s
Ø 361 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
NVIDIA-Nemotron-3-Nano-4BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181607-30d864
1.188,7 tok/s
NVIDIA-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 GeForce RTX 5090 · run-20260729-032121-058a31
1.113,5 tok/s
NVIDIA-Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-836a91
1.090,6 tok/s
NVIDIA-Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260727-181606-8b381e
1.077,4 tok/s
NVIDIA-Nemotron-3-Nano-4B3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition · run-20260728-194136-b8a818
1.071,0 tok/s
NVIDIA-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
NVIDIA-Nemotron-3-Nano-30B-A3BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260729-032121-305ad6
1.007,8 tok/s
Nemotron-Cascade-2-30B-A3BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260729-032120-4b8514
1.007,3 tok/s
Qwen3-VL-30B-A3B-Instruct this runNVIDIA GeForce RTX 5070 Ti · run-20260729-032130-26dd34
116,6 tok/s

How does this benchmark compare on other GPUs?

Same model on different hardware · 5× 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: llama.cpp
Quantization: Q4_K_M
Model: Qwen3-VL-30B-A3B-Instruct

Configuration

benchmark-konfiguration — run-20260729-032130-26dd34
# LLM-Benchmark Konfiguration # Modell : Qwen3-VL-30B-A3B-Instruct # Engine : llama.cpp # Run-ID : run-20260729-032130-26dd34 # GPU : NVIDIA GeForce RTX 5070 Ti # CPU : AMD Ryzen Threadripper PRO 5975WX 32-Cores # RAM : 247 GB bench@llm-benchmark:~$ /home/godcore/llama.cpp/build/bin/llama-server \ -m /home/godcore/.cache/huggingface/hub/models--noctrex--Huihui-Qwen3-VL-30B-A3B-Instruct-abliterated-GGUF/snapshots/6f9740ab52d5cefac63d89b65de95c12b48d49a6/Huihui-Qwen3-VL-30B-A3B-Instruct-abliterated-Q4_K_M.gguf \ --alias Qwen3-VL-30B-A3B-Instruct \ --host 0.0.0.0 \ --port 8000 \ -ngl 30 \ -c 16384 \ -np 4 \ --jinja
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-VL-30B-A3B-Instruct
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.16384
Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird./home/godcore/.cache/huggingface/hub/models--noctrex--Huihui-Qwen3-VL-30B-A3B-Instruct-abliterated-GGUF/snapshots/6f9740ab52d5cefac63d89b65de95c12b48d49a6/Huihui-Qwen3-VL-30B-A3B-Instruct-abliterated-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.30
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.16384
np4

All benchmarks of this model To leaderboard

Anzeige
Model comparison

Qwen3-VL-30B-A3B-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

2.1101.5831.0555280,006.96613.93220.898Prefill (tok/s)Generation (tok/s)NVIDIA RTX PRO 6000 Blackwell Workstation Edition - 1.639,0 tok/s Generation, 17.025 tok/s Prefill, TTFT 1.966 ms (6 Laufe)NVIDIA RTX PRO 6000 B...AMD Radeon AI PRO R9700 - 186,9 tok/s Generation, 5.398 tok/s Prefill, TTFT 1.458 ms (7 Laufe)AMD Radeon AI PRO R97...AMD Radeon 8060S Graphics - 102,2 tok/s Generation, 1.737 tok/s Prefill, TTFT 7.134 ms (3 Laufe)AMD Radeon 8060S Grap...NVIDIA GeForce RTX 5070 Ti - 154,0 tok/s Generation, 1.182 tok/s Prefill, TTFT 22.532 ms (6 Laufe) | DIESER LAUF★ NVIDIA GeForce RTX 50...
NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.639,0 tok/sAMD Radeon AI PRO R9700 186,9 tok/s★ NVIDIA GeForce RTX 5070 Ti 154,0 tok/s this runAMD Radeon 8060S Graphics 102,2 tok/s

CPUby processor

2.1101.5831.0555280,006.96613.93220.898Prefill (tok/s)Generation (tok/s)AMD Ryzen 9 9950X 16-Core Processor - 1.639,0 tok/s Generation, 17.025 tok/s Prefill, TTFT 1.966 ms (6 Laufe)AMD Ryzen 9 9950X 16-...AMD Ryzen Threadripper PRO 7955WX 16-Cores - 186,9 tok/s Generation, 5.398 tok/s Prefill, TTFT 1.458 ms (7 Laufe)AMD Ryzen Threadrippe...AMD RYZEN AI MAX+ 395 w/ Radeon 8060S - 102,2 tok/s Generation, 1.737 tok/s Prefill, TTFT 7.134 ms (3 Laufe)AMD RYZEN AI MAX+ 395...AMD Ryzen Threadripper PRO 5975WX 32-Cores - 154,0 tok/s Generation, 1.182 tok/s Prefill, TTFT 22.532 ms (6 Laufe) | DIESER LAUF★ AMD Ryzen Threadrippe...
AMD Ryzen 9 9950X 16-Core Processor 1.639,0 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 186,9 tok/s★ AMD Ryzen Threadripper PRO 5975WX 32-Cores 154,0 tok/s this runAMD RYZEN AI MAX+ 395 w/ Radeon 8060S 102,2 tok/s

MBby mainboard

2.1101.5831.0555280,006.96613.93220.898Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI - 1.639,0 tok/s Generation, 17.025 tok/s Prefill, TTFT 1.966 ms (6 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE - 186,9 tok/s Generation, 5.398 tok/s Prefill, TTFT 1.458 ms (7 Laufe)ASUSTeK COMPUTER INC....Bosgame AXB35-02 (BeyondMax Series) - 102,2 tok/s Generation, 1.737 tok/s Prefill, TTFT 7.134 ms (3 Laufe)Bosgame AXB35-02 (Bey...ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 154,0 tok/s Generation, 1.182 tok/s Prefill, TTFT 22.532 ms (6 Laufe) | DIESER LAUF★ ASUSTeK COMPUTER INC....
ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 1.639,0 tok/sASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 186,9 tok/s★ ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 154,0 tok/s this runBosgame AXB35-02 (BeyondMax Series) 102,2 tok/s

ENGby engine

1.9251.6291.3341.0387433.0256.1659.30512.445Prefill (tok/s)Generation (tok/s)vLLM - 1.028,5 tok/s Generation, 10.725 tok/s Prefill, TTFT 4.640 ms (8 Laufe)vLLMllama.cpp - 1.639,0 tok/s Generation, 4.745 tok/s Prefill, TTFT 10.105 ms (14 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 1.639,0 tok/s this runvLLM 1.028,5 tok/s

DRVby driver

1.8031.7211.6391.5571.4756.5056.7817.0587.335Prefill (tok/s)Generation (tok/s)unbekannt - 1.639,0 tok/s Generation, 6.920 tok/s Prefill, TTFT 8.118 ms (22 Laufe)unbekannt
unbekannt 1.639,0 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 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 0.22
Token / kWh1.35M
Acquisition (system)EUR 2,126 missingRAM EUR 1,976 · PSU EUR 150
Electricity (2 years)
TCO (2 years)EUR 3,755
Output tokens (2 years)7.35B
☁️ 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

Qwen3-VL-30B-A3B-InstructNVIDIA GeForce RTX 5070 TiQwen3-VL-30B-A3B-InstructNVIDIA RTX PRO 6000 Blackwell Workstation EditionQwen3-VL-30B-A3B-InstructNVIDIA RTX PRO 6000 Blackwell Workstation EditionQwen3-VL-30B-A3B-Instruct3x AMD Radeon AI PRO R9700
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.