Hey Product Hunt 👋 I'm Krish, a high school student who likes messing around with local AI and building random things that solve problems I run into. OpenBench started because I kept wanting to compare local LLMs, but doing it meant having Ollama, a terminal, GPU monitoring, and a spreadsheet open at the same time. So I built the tool I wanted. OpenBench lets you run different workloads against local models while tracking things like token speed, TTFT, VRAM usage, and GPU temperature. I've also been working on context testing, custom datasets, and LLM-as-a-judge so you can compare more than just raw tokens/sec. It's built with Rust, Tauri, and React, and it's open source. This is my first version, so I'm not pretending I've figured out the perfect way to benchmark local models yet. I'm mostly hoping people will try it, tell me what's useful, tell me what's pointless, and give me ideas for what I should build next. If you use local LLMs, I'd love to hear what you think.

OpenBench

Benchmark local LLMs without living in the terminal.

Benchmark local LLMs without living in the terminal.