Contributed byMario AlkainclusionAI Ling-3.0-flash
Performance benchmark · measured on 25.09.2026 12:53
Benchmark-ID
run-20260925-112847-06468aTimebench 3 - Kombi (Prefill + Generation)MoE127.5BRuntime: llama.cppQuantisierung: Q4_K_M
Generation11,18tok/s
Prefill107,52tok/s
Time to First Token23.546,00ms
Total duration183,93s
Concurrency1parallel
Ranking in the field
112of 122 systems
Performance benchmark · Primary metric: Generation-Speed (tok/s) · 1× concurrent
This run is better than 8 % of all comparable systems.
Generation
11,2 tok/s
-84 % vs Ø 68,2
Prefill
107,5 tok/s
-97 % vs Ø 3.254,3
Time to First Token
23.546 ms
+419 % vs Ø 4.539
Distribution in the field1 – 244 tok/s
Wie schlägt sich dieser Benchmark mit anderen Modellen?
Qwen3-0.6BAMD Radeon RX 7900 XTX · run-20260922-095735-545662
243,7 tok/s
Llama-3.2-1B-Instruct2× 2x AMD Radeon RX 7900 XTX · run-20260927-001710-d4b69b
209,0 tok/s
Llama-3.2-1B-Instruct2× 2x AMD Radeon RX 7900 XTX · run-20261008-105916-ae59c6
207,1 tok/s
gpt-oss-20bAMD Radeon RX 7900 XTX · run-20260922-115731-433d6d
154,9 tok/s
gpt-oss-20bAMD Radeon RX 7900 XTX · run-20260922-115731-0214a1
150,5 tok/s
Qwen3-0.6B2× 2x AMD Radeon RX 7900 XTX · run-20261008-105914-0699ab
143,4 tok/s
Llama-3.1-Nemotron-Nano-4B-v1.1AMD Radeon RX 7900 XTX · run-20260922-163818-53557b
139,3 tok/s
Qwen3-1.7B2× 2x AMD Radeon RX 7900 XTX · run-20261008-105915-12cae0
136,1 tok/s
gemma-4-E2B-itAMD Radeon RX 7900 XTX · run-20260922-115730-6b8875
133,2 tok/s
Qwen3-1.7B2× 2x AMD Radeon RX 7900 XTX · run-20260926-214312-91d307
132,3 tok/s
Qwen3-Coder-30B-A3B-InstructAMD Radeon RX 7900 XTX · run-20260922-140931-cdaee1
117,3 tok/s
Qwen3-30B-A3B-Instruct-2507AMD Radeon RX 7900 XTX · run-20260922-133638-f3fdcb
115,0 tok/s
Qwen3-30B-A3B-Thinking-2507AMD Radeon RX 7900 XTX · run-20260922-133638-425bd9
114,7 tok/s
Qwen3-Omni-30B-A3B-ThinkingAMD Radeon RX 7900 XTX · run-20260922-140931-defe2e
112,9 tok/s
Ling-3.0-flash this run2× 2x AMD Radeon RX 7900 XTX · run-20260925-112847-06468a
11,2 tok/s
How does this benchmark compare on other GPUs?
Same model on different hardware · 1× concurrent · Generation (tok/s)
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editionrun-20261024-140625-6a7fec
127,3 tok/s
2× NVIDIA RTX A6000run-20260930-145235-6da345
99,6 tok/s
2× NVIDIA RTX A6000run-20261001-120303-7eca8b
90,3 tok/s
2× NVIDIA RTX A6000run-20261001-081040-7e1ebf
89,7 tok/s
2× NVIDIA RTX A6000run-20260925-162606-6ab70b
87,7 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260820-050509-57046c
49,8 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260820-053757-11ac09
49,7 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260819-135555-a54bcc
49,6 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260819-192110-7d79cf
49,6 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260819-192110-0d17c1
49,6 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260820-035952-4e7bc5
49,6 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260819-195405-d529cb
49,4 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260819-195406-7021f8
49,4 tok/s
2× AMD Radeon PRO W7900 Dual Slotrun-20260819-195406-4ce3f9
49,4 tok/s
Hardware
GPU: 2x AMD Radeon RX 7900 XTX · 24 GB VRAM
CPU: AMD Ryzen Threadripper PRO 3955WX 16-Cores
RAM: 63 GB
Mainboard: ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI
Setup
Runtime: llama.cpp
Quantization: Q4_K_M
Model: Ling-3.0-flash
Configuration
# LLM-Benchmark Konfiguration
# Modell : Ling-3.0-flash
# Engine : llama.cpp
# Run-ID : run-20260925-112847-06468a
# GPU : 2x AMD Radeon RX 7900 XTX
# CPU : AMD Ryzen Threadripper PRO 3955WX 16-Cores
# RAM : 63 GB
bench@llm-benchmark:~$ llama-server \
-m ling-3.0-flash-Q4_K_M.gguf \
--alias Ling-3.0-flash \
-ngl 99 \
--n-cpu-moe 32 \
-fa on \
-c 8192 \
-np 1 '(ai-003' 2x RX 7900 XTX 'ROCm)'
| 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. | Ling-3.0-flash |
| Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann. | 8192 |
| Modellpfad?Pfad zur GGUF-Modelldatei, die geladen und ausgeliefert wird. | ling-3.0-flash-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. | 99 |
| n-cpu-moe | 32 |
| fa | on |
| Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells. | 8192 |
| np | 1 |
| execution_type | local |
Anzeige
Model comparison
Ling-3.0-flash 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
CPUby processor
AMD Ryzen 9 7945HX with Radeon Graphics 697,4 tok/sAMD EPYC 7203P 8-Core Processor 388,1 tok/sNVIDIA Grace 253,7 tok/sAMD Ryzen Threadripper PRO 5975WX 32-Cores 165,7 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 160,2 tok/s★ AMD Ryzen Threadripper PRO 3955WX 16-Cores 133,0 tok/s this runIntel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz 14,9 tok/s
MBby mainboard
Shenzhen Meigao Electronic Equipment Co.,Ltd DRFXI (MotherBoard Series) 697,4 tok/sASRockRack ROMED8-2T/BCM 388,1 tok/sASUSTeK COMPUTER INC. GX10 253,7 tok/s★ ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 194,7 tok/s this runASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 160,2 tok/sDell Inc. PowerEdge R820 14,9 tok/s
DRVby driver
unbekannt 697,4 tok/sNVIDIA 590.48.01 / CUDA 13.1 253,7 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 %
🖥️ Acquisition
EUR
🔌 Idle
10 W
⚡ TDP
720 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)720 W estimated (TDP)GPU 710 + Board 10 W full load
Avg cost / hourEUR 0.22
Electricity / 1M tokensEUR 5.37
Token / kWh55.90K
Acquisition (system)EUR 2,782 partial priceGPU EUR 2,098 · RAM EUR 504 · PSU EUR 180
Electricity (2 years)–
TCO (2 years)EUR 6,566
Output tokens (2 years)705.14M
☁️ 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
| Ling-3.0-flash2x AMD Radeon RX 7900 XTX | Ling-3.0-flashNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | Ling-3.0-flash2x NVIDIA RTX A6000 | Ling-3.0-flash2x NVIDIA 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) | – | – | – | – |
2× AMD Radeon RX 7900 XTX · this run 11 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 127 tok/s2× NVIDIA RTX A6000 100 tok/s2× NVIDIA RTX A6000 90 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.