Created bymario-alka.dePowered bygodcore.denoob2claw.detricoma.de
Contributed byMario AlkaMistral AI

Devstral-Small-2-24B-Instruct-2512

Performance benchmark · measured on 21.07.2026 16:09

Benchmark-IDrun-20260722-165907-a7f36d
Dense24BRuntime: godclawQuantisierung: Q4_K_M
Generation28,24tok/s
Prefill1.569,28tok/s
Time to First Token1.947,50ms
Total duration75,00s
Concurrency1parallel
Ranking in the field
229of 455 systems

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

This run is better than 50 % of all comparable systems.
Generation 28,2 tok/s
-20 % vs Ø 35,3
Prefill 1.569,3 tok/s
-20 % vs Ø 1.966,4
Time to First Token 1.948 ms
-95 % vs Ø 35.848
Distribution in the field1 – 148 tok/s
Ø 35 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: 3x AMD Radeon AI PRO R9700 · 32 GB VRAM
CPU: 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores
RAM: 184 GB
Mainboard: ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE

Setup

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-2-24B-Instruct-2512

Anmerkung

llama.cpp . 3 GPU . GGUF (unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF)

Configuration

benchmark-konfiguration — run-20260722-165907-a7f36d
# LLM-Benchmark Konfiguration # Modell : Devstral-Small-2-24B-Instruct-2512 # Run-ID : run-20260722-165907-a7f36d # 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-2-24B-Instruct-2512-GGUF/snapshots/6e458b8add42681bfd023de5eab93637694aaf82/Devstral-Small-2-24B-Instruct-2512-Q4_K_M.gguf Engine llamacpp Modellalias Devstral-Small-2-24B-Instruct-2512 Kontextlaenge 8192 GPU-Layer 999 Split-Mode layer 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-2-24B-Instruct-2512
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-2-24B-Instruct-2512-GGUF/snapshots/6e458b8add42681bfd023de5eab93637694aaf82/Devstral-Small-2-24B-Instruct-2512-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-2-24B-Instruct-2512
Split-Mode?Verteilung ueber mehrere GPUs: none (nur eine GPU), layer (Layer aufteilen) oder row (Tensoren zeilenweise).layer
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

Anzeige
Model comparison

Devstral-Small-2-24B-Instruct-2512 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

9417064712350,003.3986.79610.193Prefill (tok/s)Generation (tok/s)NVIDIA RTX PRO 6000 Blackwell Workstation Edition - 725,6 tok/s Generation, 8.070 tok/s Prefill, TTFT 3.939 ms (6 Laufe)NVIDIA RTX PRO 6000 B...NVIDIA GeForce RTX 5090 - 620,3 tok/s Generation, 6.308 tok/s Prefill, TTFT 6.898 ms (3 Laufe)NVIDIA GeForce RTX 50...NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition - 569,7 tok/s Generation, 8.224 tok/s Prefill, TTFT 7.901 ms (9 Laufe)NVIDIA RTX PRO 6000 B...NVIDIA GeForce RTX 3090 Ti - 331,6 tok/s Generation, 4.022 tok/s Prefill, TTFT 10.336 ms (5 Laufe)NVIDIA GeForce RTX 30...AMD Radeon PRO W7900 Dual Slot - 152,1 tok/s Generation, 1.715 tok/s Prefill, TTFT 7.815 ms (13 Laufe)AMD Radeon PRO W7900 ...NVIDIA GeForce RTX 5070 Ti - 108,5 tok/s Generation, 1.619 tok/s Prefill, TTFT 37.385 ms (3 Laufe)NVIDIA GeForce RTX 50...NVIDIA Tesla P100 PCIe 16GB - 23,3 tok/s Generation, 235 tok/s Prefill, TTFT 34.419 ms (2 Laufe)NVIDIA Tesla P100 PCI...AMD Radeon 8060S Graphics - 8,9 tok/s Generation, 22 tok/s Prefill, TTFT 906 ms (2 Laufe)AMD Radeon 8060S Grap...AMD Radeon AI PRO R9700 - 206,3 tok/s Generation, 2.714 tok/s Prefill, TTFT 13.109 ms (15 Laufe) | DIESER LAUF★ AMD Radeon AI PRO R97...
NVIDIA RTX PRO 6000 Blackwell Workstation Edition 725,6 tok/sNVIDIA GeForce RTX 5090 620,3 tok/sNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 569,7 tok/sNVIDIA GeForce RTX 3090 Ti 331,6 tok/s★ AMD Radeon AI PRO R9700 206,3 tok/s this runAMD Radeon PRO W7900 Dual Slot 152,1 tok/sNVIDIA GeForce RTX 5070 Ti 108,5 tok/sNVIDIA Tesla P100 PCIe 16GB 23,3 tok/sAMD Radeon 8060S Graphics 8,9 tok/s

CPUby processor

9417064712350,003.3986.79610.193Prefill (tok/s)Generation (tok/s)AMD Ryzen 9 9950X 16-Core Processor - 725,6 tok/s Generation, 8.070 tok/s Prefill, TTFT 3.939 ms (6 Laufe)AMD Ryzen 9 9950X 16-...AMD Ryzen 7 5800X3D 8-Core Processor - 620,3 tok/s Generation, 6.308 tok/s Prefill, TTFT 6.898 ms (3 Laufe)AMD Ryzen 7 5800X3D 8...AMD Ryzen Threadripper PRO 9965WX 24-Cores - 569,7 tok/s Generation, 8.224 tok/s Prefill, TTFT 7.901 ms (9 Laufe)AMD Ryzen Threadrippe...AMD Ryzen 9 8945HX with Radeon Graphics - 331,6 tok/s Generation, 3.668 tok/s Prefill, TTFT 14.010 ms (3 Laufe)AMD Ryzen 9 8945HX wi...AMD Ryzen 5 5600X 6-Core Processor - 247,7 tok/s Generation, 4.553 tok/s Prefill, TTFT 4.826 ms (2 Laufe)AMD Ryzen 5 5600X 6-C...AMD Ryzen Threadripper PRO 5975WX 32-Cores - 152,1 tok/s Generation, 1.697 tok/s Prefill, TTFT 13.360 ms (16 Laufe)AMD Ryzen Threadrippe...AMD Ryzen 9 7945HX with Radeon Graphics - 23,3 tok/s Generation, 235 tok/s Prefill, TTFT 34.419 ms (2 Laufe)AMD Ryzen 9 7945HX wi...AMD RYZEN AI MAX+ 395 w/ Radeon 8060S - 8,9 tok/s Generation, 22 tok/s Prefill, TTFT 906 ms (2 Laufe)AMD RYZEN AI MAX+ 395...AMD Ryzen Threadripper PRO 7955WX 16-Cores - 206,3 tok/s Generation, 2.714 tok/s Prefill, TTFT 13.109 ms (15 Laufe) | DIESER LAUF★ AMD Ryzen Threadrippe...
AMD Ryzen 9 9950X 16-Core Processor 725,6 tok/sAMD Ryzen 7 5800X3D 8-Core Processor 620,3 tok/sAMD Ryzen Threadripper PRO 9965WX 24-Cores 569,7 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 331,6 tok/sAMD Ryzen 5 5600X 6-Core Processor 247,7 tok/s★ AMD Ryzen Threadripper PRO 7955WX 16-Cores 206,3 tok/s this runAMD Ryzen Threadripper PRO 5975WX 32-Cores 152,1 tok/sAMD Ryzen 9 7945HX with Radeon Graphics 23,3 tok/sAMD RYZEN AI MAX+ 395 w/ Radeon 8060S 8,9 tok/s

MBby mainboard

9417064712350,003.3346.66810.003Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI - 725,6 tok/s Generation, 8.070 tok/s Prefill, TTFT 3.939 ms (6 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING - 620,3 tok/s Generation, 6.308 tok/s Prefill, TTFT 6.898 ms (3 Laufe)ASUSTeK COMPUTER INC....Meigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) - 331,6 tok/s Generation, 3.668 tok/s Prefill, TTFT 14.010 ms (3 Laufe)Meigao Innovation Tec...ASUSTeK COMPUTER INC. PRIME A520M-K - 247,7 tok/s Generation, 4.553 tok/s Prefill, TTFT 4.826 ms (2 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 152,1 tok/s Generation, 1.697 tok/s Prefill, TTFT 13.360 ms (16 Laufe)ASUSTeK COMPUTER INC....Shenzhen Meigao Electronic Equipment Co.,Ltd F1FXM (DeskMini Series) - 23,3 tok/s Generation, 235 tok/s Prefill, TTFT 34.419 ms (2 Laufe)Shenzhen Meigao Elect...Meigao Innovation Technology (Shen Zhen) Co., Ltd SHWSA (MS-S1 MAX) - 8,9 tok/s Generation, 22 tok/s Prefill, TTFT 906 ms (1 Lauf)Meigao Innovation Tec...Bosgame AXB35-02 (BeyondMax Series) - 8,9 tok/s Generation, 22 tok/s Prefill, TTFT 906 ms (1 Lauf)Bosgame AXB35-02 (Bey...ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE - 569,7 tok/s Generation, 4.780 tok/s Prefill, TTFT 11.156 ms (24 Laufe) | DIESER LAUF★ ASUSTeK COMPUTER INC....
ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 725,6 tok/sASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 620,3 tok/s★ ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 569,7 tok/s this runMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 331,6 tok/sASUSTeK COMPUTER INC. PRIME A520M-K 247,7 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 152,1 tok/sShenzhen Meigao Electronic Equipment Co.,Ltd F1FXM (DeskMini Series) 23,3 tok/sMeigao Innovation Technology (Shen Zhen) Co., Ltd SHWSA (MS-S1 MAX) 8,9 tok/sBosgame AXB35-02 (BeyondMax Series) 8,9 tok/s

ENGby engine

89469048728379,67223.6006.4779.354Prefill (tok/s)Generation (tok/s)vLLM - 725,6 tok/s Generation, 7.865 tok/s Prefill, TTFT 1.602 ms (3 Laufe)vLLMllama.cpp - 645,1 tok/s Generation, 3.938 tok/s Prefill, TTFT 12.776 ms (49 Laufe)llama.cppunbekannt - 247,7 tok/s Generation, 2.212 tok/s Prefill, TTFT 2.889 ms (6 Laufe)unbekannt
vLLM 725,6 tok/sllama.cpp 645,1 tok/sunbekannt 247,7 tok/s

DRVby driver

9417064712350,001.7263.4525.178Prefill (tok/s)Generation (tok/s)unbekannt - 725,6 tok/s Generation, 4.179 tok/s Prefill, TTFT 11.860 ms (54 Laufe)unbekanntROCm 7.2.0 - 8,9 tok/s Generation, 22 tok/s Prefill, TTFT 906 ms (1 Lauf)ROCm 7.2.0AMD 7.0.0-27-generic - 29,4 tok/s Generation, 1.380 tok/s Prefill, TTFT 2.259 ms (3 Laufe) | DIESER LAUF★ AMD 7.0.0-27-generic
unbekannt 725,6 tok/s★ AMD 7.0.0-27-generic 29,4 tok/s this runROCm 7.2.0 8,9 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 125 W
⚡ TDP 971 W
☁️ External LLM (API)
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 2.87
Token / kWh104.70K
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)1.78B
☁️ 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-2-24B-Instruct-25123x AMD Radeon AI PRO R9700Devstral-Small-2-24B-Instruct-2512NVIDIA RTX PRO 6000 Blackwell Workstation EditionDevstral-Small-2-24B-Instruct-2512NVIDIA GeForce RTX 5090Devstral-Small-2-24B-Instruct-2512NVIDIA RTX PRO 6000 Blackwell 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)

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.