Contributed byMario AlkaMiniMax MiniMax-M2.7
Performance benchmark · measured on 05.08.2026 22:43
Benchmark-ID
run-20260806-094439-50d485Timebench 3 - Kombi (Prefill + Generation)MoE230BRuntime: llama.cppQuantisierung: UD-Q4_K_M
Generation3,34tok/s
Prefill43,83tok/s
Time to First Token52.253,00ms
Total duration719,34s
Concurrency1parallel
Ranking in the field
60of 76 systems
Performance benchmark · Primary metric: Generation-Speed (tok/s) · 1× concurrent
This run is better than 20 % of all comparable systems.
Generation
3,3 tok/s
-98 % vs Ø 137,9
Prefill
43,8 tok/s
-98 % vs Ø 2.638,4
Time to First Token
52.253 ms
+45 % vs Ø 36.131
Distribution in the field0 – 405 tok/s
Wie schlägt sich dieser Benchmark mit anderen Modellen?
gpt-oss-20bNVIDIA GeForce RTX 5090 · run-20260724-004608-16af9b
404,6 tok/s
NVIDIA-Nemotron-3-Nano-4BNVIDIA GeForce RTX 5090 · run-20260728-194135-d456e7
393,5 tok/s
gpt-oss-20bNVIDIA GeForce RTX 5090 · run-20260724-004607-29bd16
388,9 tok/s
gpt-oss-20bNVIDIA GeForce RTX 5090 · run-20260729-032121-1c779f
388,2 tok/s
gemma-4-E2B-itNVIDIA GeForce RTX 5090 · run-20260728-184455-e8b129
377,6 tok/s
NVIDIA-Nemotron-3-Nano-30B-A3BNVIDIA GeForce RTX 5090 · run-20260729-105335-653788
359,4 tok/s
Nemotron-3-Nano-Omni-30B-A3B-ReasoningNVIDIA GeForce RTX 5090 · run-20260729-105334-2c4905
359,3 tok/s
Nemotron-Cascade-2-30B-A3BNVIDIA GeForce RTX 5090 · run-20260729-105334-44d1e7
359,0 tok/s
Qwen3-30B-A3B-Thinking-2507NVIDIA GeForce RTX 5090 · run-20260729-105335-81085a
310,1 tok/s
Qwen3-Coder-30B-A3B-InstructNVIDIA GeForce RTX 5090 · run-20260729-105335-497cfe
309,3 tok/s
Qwen3-30B-A3B-Instruct-2507NVIDIA GeForce RTX 5090 · run-20260729-105335-a4aa50
308,5 tok/s
Laguna-XS-2.1NVIDIA GeForce RTX 5090 · run-20260729-032122-816f0e
308,3 tok/s
Laguna-XS-2.1NVIDIA GeForce RTX 5090 · run-20260729-032122-45ddd9
308,2 tok/s
Laguna-XS-2.1NVIDIA GeForce RTX 5090 · run-20260729-032123-072f17
307,9 tok/s
MiniMax-M2.7 this runNVIDIA GeForce RTX 5090 · run-20260806-094439-50d485
3,3 tok/s
How does this benchmark compare on other GPUs?
Same model on different hardware · 1× concurrent · Generation (tok/s)
3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editionrun-20260806-094434-358a5d
112,8 tok/s
3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editionrun-20260806-094434-e88b45
112,6 tok/s
NVIDIA GB10run-20260723-055735-60de93
31,5 tok/s
3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Editionrun-20260806-094434-99fc33
15,1 tok/s
AMD Radeon AI PRO R9700run-20260806-094438-f664b5
9,0 tok/s
AMD Radeon AI PRO R9700run-20260806-094438-cfda1f
9,0 tok/s
AMD Radeon AI PRO R9700run-20260806-094438-e40031
9,0 tok/s
3× AMD Radeon AI PRO R9700run-20260726-084121-1490ea
7,1 tok/s
NVIDIA RTX PRO 6000 Blackwell Workstation Editionrun-20260806-094439-70a150
5,2 tok/s
NVIDIA GeForce RTX 3090 Tirun-20260730-120419-2ad3a7
4,0 tok/s
NVIDIA GeForce RTX 5090 this runrun-20260806-094439-50d485
3,3 tok/s
Keine GPU (CPU-only)run-20260723-114454-a888a9
2,6 tok/s
Hardware
GPU: NVIDIA GeForce RTX 5090 · 32 GB VRAM
CPU: AMD Ryzen 7 5800X3D 8-Core Processor
RAM: 126 GB
Mainboard: ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING
Setup
Runtime: llama.cpp
Quantization: UD-Q4_K_M
Model: MiniMax-M2.7
Configuration
# LLM-Benchmark Konfiguration
# Modell : MiniMax-M2.7
# Engine : llama.cpp
# Run-ID : run-20260806-094439-50d485
# GPU : NVIDIA GeForce RTX 5090
# CPU : AMD Ryzen 7 5800X3D 8-Core Processor
# RAM : 126 GB
bench@llm-benchmark:~$ /root/llama.cpp/build/bin/llama-server \
-m /root/.cache/huggingface/hub/models--unsloth--MiniMax-M2.7-GGUF/snapshots/d2a05ccf69491b03db0cc40b335aec14bdaf7198/UD-Q4_K_M/MiniMax-M2.7-UD-Q4_K_M-00001-of-00004.gguf \
--alias MiniMax-M2.7 \
--host 0.0.0.0 \
--port 8000 \
-ngl 12 \
-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. | MiniMax-M2.7 |
| 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. | /root/.cache/huggingface/hub/models--unsloth--MiniMax-M2.7-GGUF/snapshots/d2a05ccf69491b03db0cc40b335aec14bdaf7198/UD-Q4_K_M/MiniMax-M2.7-UD-Q4_K_M-00001-of-00004.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. | 12 |
| Kontext?Groesse des Kontextfensters in Token. 0 = der beim Training verwendete Kontext des Modells. | 16384 |
| np | 4 |
Anzeige
Model comparison
MiniMax-M2.7 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 Threadripper PRO 9965WX 24-Cores 548,6 tok/sNVIDIA Grace 174,2 tok/sAMD Ryzen Threadripper PRO 7955WX 16-Cores 16,9 tok/sAMD Ryzen 9 9950X 16-Core Processor 7,0 tok/s★ AMD Ryzen 7 5800X3D 8-Core Processor 6,0 tok/s this runAMD Ryzen 9 8945HX with Radeon Graphics 4,0 tok/sIntel(R) Xeon(R) CPU E5-4620 v2 @ 2.60GHz 2,6 tok/s
MBby mainboard
ASUSTeK COMPUTER INC. Pro WS WRX90E-SAGE SE 548,6 tok/sASUSTeK COMPUTER INC. GX10 174,2 tok/sASUSTeK COMPUTER INC. ProArt X870E-CREATOR WIFI 7,0 tok/s★ ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 6,0 tok/s this runMeigao Innovation Technology (Shen Zhen) Co., Ltd DRFXL (MotherBoard Series) 4,0 tok/sDell Inc. PowerEdge R820 2,6 tok/s
DRVby driver
unbekannt 548,6 tok/sNVIDIA 590.48.01 / CUDA 13.1 174,2 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
72 W
⚡ TDP
622 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)622 W estimated (TDP)GPU 575 + CPU 35 + Board 12 W full load
Avg cost / hourEUR 0.19
Electricity / 1M tokensEUR 15.51
Token / kWh19.35K
Acquisition (system)EUR 4,986 full priceGPU EUR 3,300 · CPU EUR 349 · Board EUR 149 · RAM EUR 1,008 · PSU EUR 180
Electricity (2 years)–
TCO (2 years)EUR 8,253
Output tokens (2 years)210.66M
☁️ 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 (72 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
| MiniMax-M2.7NVIDIA GeForce RTX 5090 | MiniMax-M2.73x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | MiniMax-M2.73x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | MiniMax-M2.7NVIDIA GB10 (DGX Spark) | |
|---|---|---|---|---|
| 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) | – | – | – | – |
NVIDIA GeForce RTX 5090 · this run 3 tok/s3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 113 tok/s3× NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 113 tok/sNVIDIA GB10 (DGX Spark) 32 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.