benchbro

benchmark your local LLMs without the headache. runs in a browser, talks to ollama or llama.cpp.

early/alpha - works but is quite rough around the edges.

setup

pip install -e .
cd frontend && npm install && npm run build && cd ..

usage

# have ollama running with a model pulled, then:
benchbro

opens a browser. pick your model, pick a benchmark, run it, see results.

benchmarks

• MMLU-Pro (knowledge)
• GSM8K (math)
• HumanEval (code)
• Perplexity (quant quality)

license

MIT