Free · Open source · Works with Ollama & LM Studio

Which LLM runs best on YOUR machine?

Stop guessing. Benchmark local LLMs directly on your hardware — speed, quality, memory — and get a clear fitness verdict.

$ npm install -g metrillm@latest
$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest

One command. Real answers.

llama3.2:3b Great Fit

42.1

tok/s

68%

quality

3.2 GB

memory

MacBook Air M2 · 16 GB

qwen2.5:7b Good Fit

28.5

tok/s

74%

quality

5.8 GB

memory

Mac Mini M4 Pro · 48 GB

gemma3:12b Moderate

14.3

tok/s

81%

quality

9.4 GB

memory

Desktop RTX 4090 · 64 GB

Example results — yours will reflect your actual hardware.

How it works

1

Install & run

One command. No config. Works with Ollama and LM Studio.

2

Get your verdict

Speed, quality, memory — tested with 14 targeted prompts.

3

Join the leaderboard

Upload your results. Compare with real hardware from the community.

The community is benchmarking

llama3.2qwen2.5gemma3phi-4mistraldeepseek-r1command-rllama3.3
Apple M2Apple M4 ProApple M4 MaxRTX 4090RTX 3080AMD Ryzen 9

Why MetriLLM

Open source

Free CLI, public methodology. Every result is reproducible.

Your hardware, your data

Benchmarks run locally. No cloud, no uploading your models.

Real-world testing

14 targeted prompts covering reasoning, math, coding, and more.

Community-driven

Real results on real hardware. No synthetic benchmarks.

Ready to find out?

Run your first benchmark. Free and open source.

$ npm install -g metrillm@latest
$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest

Don't miss new benchmarks

Get notified when new models and hardware configurations are tested. No spam, unsubscribe anytime.