I built Benchmark Registry because finding AI benchmark results is way harder and fragmented than it should be. Results are scattered across model cards, research papers, blog posts, announcements, etc. Comparing models often means hunting through all of them, and even then, details like reasoning level, benchmark version, evaluator, and source are missing. Benchmark Registry is my attempt to fix this problem. We’re launching with 500+ benchmark results across 92 AI models, with every result tied back to its source. The project is also fully open source. This is just the start. My vision is for Benchmark Registry to grow into a useful public reference for researchers, developers, and anyone trying to understand how AI models perform. I’d love some feedback on what models, benchmarks, or data sources you think we should add next. Thank you!

Benchmark Registry

One place for AI benchmark results.

One place for AI benchmark results.