Created bymario-alka.dePowered bynoob2claw.detricoma.de
Contributed byMario AlkaNVIDIA

Llama-3.1-Nemotron-Nano-4B-v1.1

Performance benchmark · measured on 08.11.2026 20:07

Benchmark-IDrun-20261108-194953-a60716
Timebench 3 - Kombi (Prefill + Generation)Dense4.5BRuntime: vLLM
Generation86,25tok/s
Prefill8.338,76tok/s
Time to First Token298,00ms
Total duration15,14s
Concurrency1parallel
Ranking in the field
642of 1927 systems

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

This run is better than 67 % of all comparable systems.
Generation 86,3 tok/s
+11 % vs Ø 77,5
Prefill 8.338,8 tok/s
+193 % vs Ø 2.844,6
Time to First Token 298 ms
-99 % vs Ø 23.755
Distribution in the field0 – 471 tok/s
Ø 78 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: NVIDIA GeForce RTX 3090 · 24 GB VRAM
CPU: AMD Ryzen 5 5600X 6-Core Processor
RAM: 31 GB
Mainboard: ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING

Setup

Runtime: vLLM
Quantization: -
Model: Llama-3.1-Nemotron-Nano-4B-v1.1

Configuration

benchmark-konfiguration — run-20261108-194953-a60716
# LLM-Benchmark Konfiguration # Modell : Llama-3.1-Nemotron-Nano-4B-v1.1 # Engine : vLLM # Run-ID : run-20261108-194953-a60716 # GPU : NVIDIA GeForce RTX 3090 # CPU : AMD Ryzen 5 5600X 6-Core Processor # RAM : 31 GB bench@llm-benchmark:~$ vllm serve \ --model Llama-3.1-Nemotron-Nano-4B-v1.1
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.vllm
Kontextlaenge?Maximale Anzahl Tokens (Eingabe + erzeugte Ausgabe zusammen), die das Modell pro Anfrage verarbeiten kann.4096
execution_typelocal

All benchmarks of this model To leaderboard

Anzeige
Model comparison

Llama-3.1-Nemotron-Nano-4B-v1.1 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

5354022681340,02363.6527.06910.486Prefill (tok/s)Generation (tok/s)AMD Radeon RX 7900 XTX - 419,4 tok/s Generation, 8.744 tok/s Prefill, TTFT 1.556 ms (6 Laufe)AMD Radeon RX 7900 XTXCPU-only - 49,2 tok/s Generation, 1.978 tok/s Prefill, TTFT 1.328 ms (1 Lauf)CPU-onlyNVIDIA GeForce RTX 3090 - 86,3 tok/s Generation, 8.339 tok/s Prefill, TTFT 298 ms (1 Lauf) | DIESER LAUF★ NVIDIA GeForce RTX 30...
AMD Radeon RX 7900 XTX 419,4 tok/s★ NVIDIA GeForce RTX 3090 86,3 tok/s this runCPU-only 49,2 tok/s

CPUby processor

5354022681340,02363.6527.06910.486Prefill (tok/s)Generation (tok/s)AMD Ryzen Threadripper PRO 3955WX 16-Cores - 419,4 tok/s Generation, 8.744 tok/s Prefill, TTFT 1.556 ms (6 Laufe)AMD Ryzen Threadrippe...Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) - 49,2 tok/s Generation, 1.978 tok/s Prefill, TTFT 1.328 ms (1 Lauf)Apple M6 (unified CPU...AMD Ryzen 5 5600X 6-Core Processor - 86,3 tok/s Generation, 8.339 tok/s Prefill, TTFT 298 ms (1 Lauf) | DIESER LAUF★ AMD Ryzen 5 5600X 6-C...
AMD Ryzen Threadripper PRO 3955WX 16-Cores 419,4 tok/s★ AMD Ryzen 5 5600X 6-Core Processor 86,3 tok/s this runApple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) 49,2 tok/s

MBby mainboard

5354022681340,02363.6527.06910.486Prefill (tok/s)Generation (tok/s)ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI - 419,4 tok/s Generation, 8.744 tok/s Prefill, TTFT 1.556 ms (6 Laufe)ASUSTeK COMPUTER INC....Apple Mac mini - 49,2 tok/s Generation, 1.978 tok/s Prefill, TTFT 1.328 ms (1 Lauf)Apple Mac miniASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING - 86,3 tok/s Generation, 8.339 tok/s Prefill, TTFT 298 ms (1 Lauf) | DIESER LAUF★ ASUSTeK COMPUTER INC....
ASUSTeK COMPUTER INC. Pro WS WRX80E-SAGE SE WIFI 419,4 tok/s★ ASUSTeK COMPUTER INC. ROG STRIX B550-A GAMING 86,3 tok/s this runApple Mac mini 49,2 tok/s

ENGby engine

5354022681340,02363.6527.06910.486Prefill (tok/s)Generation (tok/s)llama.cpp - 419,4 tok/s Generation, 8.744 tok/s Prefill, TTFT 1.556 ms (6 Laufe)llama.cpplmstudio-mlx - 49,2 tok/s Generation, 1.978 tok/s Prefill, TTFT 1.328 ms (1 Lauf)lmstudio-mlxvLLM - 86,3 tok/s Generation, 8.339 tok/s Prefill, TTFT 298 ms (1 Lauf) | DIESER LAUF★ vLLM
llama.cpp 419,4 tok/s★ vLLM 86,3 tok/s this runlmstudio-mlx 49,2 tok/s

DRVby driver

4614404193983777.3777.6908.0048.318Prefill (tok/s)Generation (tok/s)unbekannt - 419,4 tok/s Generation, 7.847 tok/s Prefill, TTFT 1.370 ms (8 Laufe)unbekannt
unbekannt 419,4 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 12 W
⚡ TDP 362 W
☁️ External LLM (API)
Electricity0.30 EUR/kWh
Avg power (incl. idle)362 W estimated (TDP)GPU 350 + Board 12 W full load
Avg cost / hourEUR 0.11
Electricity / 1M tokensEUR 0.35
Token / kWh857.73K
Acquisition (system)EUR 1,518 full priceGPU EUR 850 · CPU EUR 121 · Board EUR 149 · RAM EUR 248 · PSU EUR 150
Electricity (2 years)–
TCO (2 years)EUR 3,421
Output tokens (2 years)5.44B
☁️ 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 (12 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

Llama-3.1-Nemotron-Nano-4B-v1.1NVIDIA GeForce RTX 3090Llama-3.1-Nemotron-Nano-4B-v1.1AMD Radeon RX 7900 XTXLlama-3.1-Nemotron-Nano-4B-v1.1AMD Radeon RX 7900 XTXLlama-3.1-Nemotron-Nano-4B-v1.1Keine GPU (CPU-only)
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.