# MetriLLM

> Open-source CLI tool and leaderboard for benchmarking local LLM inference on real hardware. Measures speed (tokens/s, TTFT), memory usage, and response quality (reasoning, math, coding, instruction following, structured output, multilingual). MIT license, npm package.

- [Homepage](https://metrillm.dev/): Live leaderboard with benchmark results across real hardware configurations
- [CLI on GitHub](https://github.com/MetriLLM/metrillm): Source code, installation instructions, and contribution guide

## Documentation

- [Methodology](https://metrillm.dev/methodology): How benchmarks are scored — global score formula, hardware fit, quality metrics, verdict system, hardware profiles
- [Specification](https://metrillm.dev/specification): Benchmark output format and data schema
- [Models directory](https://metrillm.dev/models): All benchmarked models with individual performance pages
- [Hardware directory](https://metrillm.dev/hardware): All benchmarked hardware configurations with per-CPU pages

## Community

- [Blog](https://metrillm.dev/blog): Articles about local LLM performance and benchmarking insights
- [Contributors](https://metrillm.dev/contributors): People who contributed benchmark data
- [Founding Supporters](https://metrillm.dev/founding): Early supporter program

## Tools

- [My Benchmarks](https://metrillm.dev/my-benchmarks): Look up your own benchmark results by machine ID
- [Compare](https://metrillm.dev/compare): Side-by-side model comparison tool

## Optional

- [Privacy Policy](https://metrillm.dev/privacy): Privacy policy
- [Terms of Service](https://metrillm.dev/terms): Terms of service
- [Legal](https://metrillm.dev/legal): Legal notices