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

command-a-reasoning-08-2025

Performance benchmark · measured on 27.07.2026 18:56

Benchmark-IDrun-20260727-193100-fec8b5
Timebench 3 - Kombi (Prefill + Generation)111BRuntime: llama.cppQuantisierung: Q4_K_M
Generation156,96tok/s
Prefill1.526,54tok/s
Time to First Token53.068,50ms
Total duration287,03s
Concurrency10parallel
Ranking in the field
18of 20 systems

Performance benchmark · Primary metric: Generation-Speed (tok/s) · 10× concurrent

This run is better than 11 % of all comparable systems.
Generation 157,0 tok/s
-83 % vs Ø 945,8
Prefill 1.526,5 tok/s
-92 % vs Ø 19.232,0
Time to First Token 53.069 ms
+132 % vs Ø 22.897
Distribution in the field0 – 2.414 tok/s
Ø 946 Median Dieser Lauf

Wie schlägt sich dieser Benchmark mit anderen Modellen?

gemma-4-E2B-itNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181606-b37ed9
2.414,4 tok/s
NVIDIA-Nemotron-3-Nano-4BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181607-4cca42
2.182,7 tok/s
gemma-4-E4B-itNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-181607-866815
1.758,6 tok/s
gpt-oss-20bNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260723-194447-dc199d
1.443,9 tok/s
Qwen3-Coder-30B-A3B-InstructNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-004606-ebf474
1.070,6 tok/s
Qwen3.5-35B-A3BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-082959-cd1403
1.028,6 tok/s
Qwen3-VL-30B-A3B-InstructNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-004607-c49303
1.028,5 tok/s
Ministral-3-14B-Reasoning-2512NVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260723-195749-eebab2
1.015,7 tok/s
Qwen3-30B-A3B-Instruct-2507NVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-004604-860045
961,5 tok/s
Qwen3.6-35B-A3BNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-083000-8690d7
952,8 tok/s
Qwen3-Omni-30B-A3B-ThinkingNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-004611-a583d6
910,7 tok/s
Qwen3-30B-A3B-Thinking-2507NVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-004605-f9a3a1
906,0 tok/s
Magistral-Small-2509NVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-004604-c6df15
726,9 tok/s
Devstral-Small-2-24B-Instruct-2512NVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260724-004603-9d3130
725,6 tok/s
command-a-reasoning-08-2025 this runNVIDIA RTX PRO 6000 Blackwell Workstation Edition · run-20260727-193100-fec8b5
157,0 tok/s

How does this benchmark compare on other GPUs?

Same model on different hardware · 10× concurrent · Generation (tok/s)

Hardware

GPU: NVIDIA RTX PRO 6000 Blackwell Workstation Edition · 96 GB VRAM
CPU: AMD Ryzen 9 9950X 16-Core Processor
RAM: 92 GB
Mainboard: ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI

Setup

Runtime: llama.cpp
Quantization: Q4_K_M
Model: command-a-reasoning-08-2025

Configuration

benchmark-konfiguration — run-20260727-193100-fec8b5
# LLM-Benchmark Konfiguration # Modell : command-a-reasoning-08-2025 # Engine : llama.cpp # Run-ID : run-20260727-193100-fec8b5 # GPU : NVIDIA RTX PRO 6000 Blackwell Workstation Edition # CPU : AMD Ryzen 9 9950X 16-Core Processor # RAM : 92 GB bench@llm-benchmark:~$ /home/godcore/llama.cpp/build/bin/llama-server \ -m /home/godcore/.cache/huggingface/hub/models--DevQuasar--CohereLabs.command-a-reasoning-08-2025-GGUF/snapshots/924408415a0e807b935d0a0c4f316a1bc40a1051/CohereLabs.command-a-reasoning-08-2025.Q4_K_M-00001-of-00005.gguf \ --alias command-a-reasoning-08-2025 \ --host 0.0.0.0 \ --port 8000 \ -ngl 999 \ -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.command-a-reasoning-08-2025
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--DevQuasar--CohereLabs.command-a-reasoning-08-2025-GGUF/snapshots/924408415a0e807b935d0a0c4f316a1bc40a1051/CohereLabs.command-a-reasoning-08-2025.Q4_K_M-00001-of-00005.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.16384
np4

All benchmarks of this model To leaderboard

Model comparison

command-a-reasoning-08-2025 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

20315210250,80,006871.3752.062Prefill (tok/s)Generation (tok/s)NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition - 134,9 tok/s Generation, 1.676 tok/s Prefill, TTFT 26.548 ms (3 Laufe)NVIDIA RTX PRO 6000 B...NVIDIA GeForce RTX 5070 Ti - 6,7 tok/s Generation, 108 tok/s Prefill, TTFT 64.666 ms (3 Laufe)NVIDIA GeForce RTX 50...NVIDIA GeForce RTX 3090 Ti - 4,2 tok/s Generation, 90 tok/s Prefill, TTFT 72.246 ms (3 Laufe)NVIDIA GeForce RTX 30...NVIDIA RTX PRO 6000 Blackwell Workstation Edition - 157,0 tok/s Generation, 1.435 tok/s Prefill, TTFT 25.033 ms (3 Laufe) | DIESER LAUF★ NVIDIA RTX PRO 6000 B...
★ NVIDIA RTX PRO 6000 Blackwell Workstation Edition 157,0 tok/s this runNVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 134,9 tok/sNVIDIA GeForce RTX 5070 Ti 6,7 tok/sNVIDIA GeForce RTX 3090 Ti 4,2 tok/s

CPUby processor

20315210250,80,006871.3752.062Prefill (tok/s)Generation (tok/s)AMD Ryzen Threadripper PRO 9965WX 24-Cores - 134,9 tok/s Generation, 1.676 tok/s Prefill, TTFT 26.548 ms (3 Laufe)AMD Ryzen Threadrippe...AMD Ryzen Threadripper PRO 5975WX 32-Cores - 6,7 tok/s Generation, 108 tok/s Prefill, TTFT 64.666 ms (3 Laufe)AMD Ryzen Threadrippe...AMD Ryzen 9 8945HX with Radeon Graphics - 4,2 tok/s Generation, 90 tok/s Prefill, TTFT 72.246 ms (3 Laufe)AMD Ryzen 9 8945HX wi...AMD Ryzen 9 9950X 16-Core Processor - 157,0 tok/s Generation, 1.435 tok/s Prefill, TTFT 25.033 ms (3 Laufe) | DIESER LAUF★ AMD Ryzen 9 9950X 16-...
★ AMD Ryzen 9 9950X 16-Core Processor 157,0 tok/s this runAMD Ryzen Threadripper PRO 9965WX 24-Cores 134,9 tok/sAMD Ryzen Threadripper PRO 5975WX 32-Cores 6,7 tok/sAMD Ryzen 9 8945HX with Radeon Graphics 4,2 tok/s

MBby mainboard

20315210250,80,006871.3752.062Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE - 134,9 tok/s Generation, 1.676 tok/s Prefill, TTFT 26.548 ms (3 Laufe)ASUSTeK COMPUTER INC....ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 6,7 tok/s Generation, 108 tok/s Prefill, TTFT 64.666 ms (3 Laufe)ASUSTeK COMPUTER INC....Meigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) - 4,2 tok/s Generation, 90 tok/s Prefill, TTFT 72.246 ms (3 Laufe)Meigao Innovation Tec...ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI - 157,0 tok/s Generation, 1.435 tok/s Prefill, TTFT 25.033 ms (3 Laufe) | DIESER LAUF★ ASUSTeK COMPUTER INC....
★ ASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 157,0 tok/s this runASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 134,9 tok/sASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 6,7 tok/sMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 4,2 tok/s

ENGby engine

173165157149141778811844877Prefill (tok/s)Generation (tok/s)llama.cpp - 157,0 tok/s Generation, 827 tok/s Prefill, TTFT 47.123 ms (12 Laufe) | DIESER LAUF★ llama.cpp
★ llama.cpp 157,0 tok/s this run

DRVby driver

173165157149141778811844877Prefill (tok/s)Generation (tok/s)unbekannt - 157,0 tok/s Generation, 827 tok/s Prefill, TTFT 47.123 ms (12 Laufe)unbekannt
unbekannt 157,0 tok/s
💰 Economics

Economics of this run

Operating cost, TCO and comparison with the next-best runs of the same model at identical concurrency (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 70 W
⚡ TDP 644 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)644 W estimated (TDP)GPU 600 + CPU 29 + Board 15 W full load
Avg cost / hourEUR 0.19
Electricity / 1M tokensEUR 0.34
Token / kWh877.76K
Acquisition (system)EUR 15,248 full priceGPU EUR 13,000 · CPU EUR 649 · Board EUR 499 · RAM EUR 920 · PSU EUR 180
Electricity (2 years)
TCO (2 years)EUR 18,632
Output tokens (2 years)9.90B
☁️ 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 (70 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

command-a-reasoning-08-2025NVIDIA RTX PRO 6000 Blackwell Workstation Editioncommand-a-reasoning-08-20253x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editioncommand-a-reasoning-08-2025NVIDIA GeForce RTX 5070 Ticommand-a-reasoning-08-2025NVIDIA GeForce RTX 3090 Ti
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.