Runtime: godclaw
Quantization: Q4_K_M
Operating system: Ubuntu 26.04 LTS (Kernel 7.0.0-27-generic)
Driver: AMD 7.0.0-27-generic
Model: Devstral-Small-2507

Anmerkung

llama.cpp . 1 GPU . GGUF (unsloth/Devstral-Small-2507-GGUF)

Configuration

benchmark-konfiguration — run-20260722-165907-bc8536

# LLM-Benchmark Konfiguration # Modell : Devstral-Small-2507 # Run-ID : run-20260722-165907-bc8536 # GPU : 3x AMD Radeon AI PRO R9700 # CPU : 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores # RAM : 184 GB bench@llm-benchmark:~$ cat benchmark.conf Konfigurationspfad /root/.cache/huggingface/hub/models--unsloth--Devstral-Small-2507-GGUF/snapshots/8d0c7eb7bf0142cc3b94c88450eddcc7720e4488/Devstral-Small-2507-Q4_K_M.gguf Engine llamacpp Modellalias Devstral-Small-2507 Kontextlaenge 8192 GPU-Layer 999 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.	llamacpp
Modellalias?Der Name, unter dem das Modell ueber die API angesprochen wird. Genau dieser Wert muss im Request-Feld 'model' stehen.	Devstral-Small-2507
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.	/root/.cache/huggingface/hub/models--unsloth--Devstral-Small-2507-GGUF/snapshots/8d0c7eb7bf0142cc3b94c88450eddcc7720e4488/Devstral-Small-2507-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.	999
Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells.	8192
Alias?Anzeigename des Modells nach aussen (served model name), unabhaengig vom Dateinamen.	Devstral-Small-2507
Host?Netzwerk-Interface, an das der HTTP-Server bindet, z.B. 0.0.0.0 fuer alle Interfaces.	0.0.0.0
Port?TCP-Port des HTTP-Servers.	8000

All benchmarks of this model To leaderboard

Model comparison

Devstral-Small-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 RTX PRO 6000 Blackwell Workstation Edition 647,6 tok/sNVIDIA GeForce RTX 5090 625,1 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 572,2 tok/sNVIDIA GeForce RTX 3090 Ti 317,4 tok/s★ AMD Radeon AI PRO R9700 205,8 tok/s this runAMD Radeon PRO W7900 Dual Slot 137,5 tok/sNVIDIA GeForce RTX 5070 Ti 103,9 tok/sNVIDIA Tesla P100 PCIe 16GB 23,2 tok/s

CPUby processor

AMD Ryzen 9 9950X 16-Core Processor 647,6 tok/sAMD Ryzen 7 5800X3D 8-Core Processor 625,1 tok/sAMD Ryzen Threadripper PRO 9965WX 24-Cores 572,2 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 317,4 tok/s★ AMD Ryzen Threadripper PRO 7955WX 16-Cores 205,8 tok/s this runAMD Ryzen Threadripper PRO 5975WX 32-Cores 137,5 tok/sAMD Ryzen 5 5600X 6-Core Processor 57,1 tok/sAMD Ryzen 9 7945HX with Radeon Graphics 23,2 tok/s

MBby mainboard

ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 647,6 tok/sASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 625,1 tok/s★ ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 572,2 tok/s this runMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 317,4 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 137,5 tok/sASUSTeK COMPUTER INC. PRIME A520M-K 57,1 tok/sShenzhen Meigao Electronic Equipment Co.,Ltd F1FXM (DeskMini Series) 23,2 tok/s

ENGby engine

llama.cpp 647,6 tok/sunbekannt 57,1 tok/s

DRVby driver

unbekannt 647,6 tok/s★ AMD 7.0.0-27-generic 31,4 tok/s this run

💰 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 URL Reset

⚡ Electricity price EUR/kWh

⚙️ System utilization 100 %

🖥️ Acquisition EUR

🔌 Idle 125 W

⚡ TDP 971 W

☁️ External LLM (API) show

Input €/1M Output €/1M Input per output token

Electricity0.30 EUR/kWh

Avg power (incl. idle)971 W estimated (TDP)GPU 900 + CPU 46 + Board 25 W full load

Avg cost / hourEUR 0.29

Electricity / 1M tokensEUR 2.84

Token / kWh105.48K

Acquisition (system)EUR 9,674 full priceGPU EUR 4,200 · CPU EUR 1,399 · Board EUR 1,299 · RAM EUR 2,576 · PSU EUR 200

Electricity (2 years)–

TCO (2 years)EUR 14,778

Output tokens (2 years)1.79B

☁️ 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 (125 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

	Devstral-Small-25073x AMD Radeon AI PRO R9700	Devstral-Small-2507NVIDIA GeForce RTX 5090	Devstral-Small-2507NVIDIA RTX PRO 6000 Blackwell Workstation Edition	Devstral-Small-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)	–	–	–	–

3× AMD Radeon AI PRO R9700 · this run 28 tok/s NVIDIA GeForce RTX 5090 98 tok/s NVIDIA RTX PRO 6000 Blackwell Workstation Edition 97 tok/s 3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 93 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.