Community benchmarks

Model Arena

See how local AI models actually perform — instruction following, tool calling, research, reasoning, and coding — benchmarked by the community and shared in one place.

Browse results

Real scores across research, tool-calling, reasoning, and coding tests — not marketing benchmarks.

Contribute your own

Run the local benchmark suite against your own models and upload just the scores — never your API keys or full transcripts.

Run it yourself

The desktop benchmark runner is $1/mo — free for bravomedia.com accounts. Sign in to download.