MetriLLM FAQs

Name: MetriLLM
Author: MetriLLM

Question 1

What is MetriLLM and what does it do?

Accepted Answer

MetriLLM is a powerful benchmarking tool designed for local Large Language Models (LLMs). It assesses an LLM's speed, quality, and compatibility with your hardware, providing a comprehensive global score and public leaderboard rank.

Question 2

How does MetriLLM evaluate LLM performance and quality?

Accepted Answer

It measures performance metrics like tokens per second, time to first token (TTFT), and memory usage. For quality, it evaluates models across 6 categories (reasoning, coding, math, instruction following, structured output, multilingual) using 14 prompts.

Question 3

Can I share my MetriLLM benchmark results?

Accepted Answer

Yes, MetriLLM allows one-click sharing of your benchmark results to a public leaderboard. This generates a unique URL for your test, displays your global rank, and even provides a shareable card for social media.

Question 4

What are MetriLLM's integration options and system requirements?

Accepted Answer

MetriLLM integrates seamlessly via a CLI, MCP servers for AI coding assistants (like Claude Code, Cursor), and dedicated IDE plugins. It requires Node 20+ and Ollama to be running on your system.

Question 5

How is the MetriLLM Global Score determined?

Accepted Answer

The Global Score (0-100) is calculated as 40% Hardware Fit and 60% Quality. Hardware Fit considers speed, TTFT, and memory efficiency, while Quality scores responses across evaluation categories. The scoring adapts to your hardware tier (Entry/Balanced/High-End).

MetriLLM

MetriLLM

主な機能

ユースケース

主な機能

ユースケース