llama3.3:70b

llama · 70.6B · Q4_K_M

NVIDIA NVIDIA_DGX_Spark (Cortex-X925)

122 GB · Ubuntu 24.04.4 LTS

Tested on June 26, 2026

Top 49% Compare

Global Score

73 /100

Not Rec.

Hardware Fit

51/100

Quality

83/100

Get this model

🦙

Ollama

ollama pull llama3.3:70b

View on Ollama Library

ollama.com/library/llama3.3

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: NVIDIA NVIDIA_DGX_Spark
CPU: Cortex-X925
Cores: 20 total (20 perf)
Frequency: 3.35 GHz
RAM: 122 GB
GPU: Device 2e12
OS: Ubuntu 24.04.4 LTS
Arch: arm64
Power Mode: performance

Performance

Tokens/sec Tokens generated per second — higher is better: 4.6
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±0.0
First chunk latency Delay before the first streamed chunk arrives from runtime: 560 ms
Time to first token How fast the model starts responding — lower is better: 560 ms
Load time Time to load the model into memory before inference: 13.8 s
Memory usage RAM consumed during inference — percentage of total system memory: 49.5 GB (41%)
Total tokens Total number of tokens generated across all benchmark prompts: 982

Score breakdown

Speed

4/50

Time to first token

20/20

Memory

27/30

Quality

Reasoning

17/20

Coding

18/20

Instruction following

12/20

Structured output

15/15

Math

11/15

Multilingual

10/10

Category levels

Reasoning: Strong Coding: Strong Instruction Following: Adequate Structured Output: Strong Math: Adequate Multilingual: Strong

Metadata

Spec version: 0.2.1
Runtime: Ollama 0.30.10
Model format: GGUF
Hardware profile: HIGH-END
Result hash: 2ceb7f05982f67ac27ec9488eaf22612d401c814ff2cd3c9bebf159647b63afe

Interpretation

Hardware fit: 51/100. Overall suitability: NOT RECOMMENDED (Global 73/100). Category profile: Reasoning: Strong, Coding: Strong, Instruction Following: Adequate, Structured Output: Strong, Math: Adequate, Multilingual: Strong.

Disqualifiers

Token speed too low: 4.6 tok/s (minimum: 7 tok/s for HIGH-END profile)

Bench Environment

Thermal: nominal Swap delta: +0.0 GB CPU load: avg 15% (peak 35%)

Run yours now

$ npm install -g metrillm@latest

$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest