Community benchmark aggregator for local LLM performance

llm-bench.io — find the right LLM
for your usecase and hardware.

I built this website because finding the best local LLM for your local setup shouldn't take weeks (like it did for me). Leverage the power of community benchmarks to find models that fit your GPU, deliver the speed you need, and provide quality output for your specific usecase — all before spending hours on research and testing. Enjoy!

View BenchmarksRun Benchmark
—
models benchmarked
—
runs within last 24h
—
total runs

Which model for your memory?

Every model the community ran on machines with this much memory: how fast it generates, and how well it answered the four scenarios. Top-left is slow and good, bottom-right fast and weak. The green ones are the best trade-offs.

Best trade-off: quality 60+, and nothing else here is both faster and betterOther models · dot size = number of runs
Loading…

Leaderboard

Loading...
View all benchmarks →

Contribute Your Results

Help others make informed decisions. Download the client, benchmark your setup, and share your results with the community.

Quick Install

Linux/macOS

curl -fsSL https://llm-bench.io/install.sh | bash

Windows

irm https://llm-bench.io/install.ps1 | iex
# run the benchmark
$ ./llm-benchmark
Full Setup Guide →

How We Test

We benchmark models on real-world scenarios: code generation, reasoning, roleplay, and research. Measuring both speed and output quality so you can make the right choice.

Read the Methodology