Welcome to the ChessBench community!

[benjaminbrumfield]

ChessBench is a new chess benchmark for language models. The site is at chessbench.ai. This is the community side.

ChessBench is also a research platform for studying what elicits a model's best play and what its games reveal about latent capabilities. If you want to dig into methodology or have ideas for experiments, the Q&A section is the place.

Things you can do here:

• Request a model. Is the leaderboard missing a model you'd like to see? Open a Model Request. Tell me what it is and how to access it.
• Request a matchup. Want to see two specific players go head-to-head? Matchup Requests. (Heads up: matchup-level data isn't surfaced on the site yet, but it's coming.)
• Suggest a feature. Something the site should do that it doesn't? Feature Ideas.
• Ask a question. Methodology, Elo math, why your favorite model lost to Stockfish-1100. Q&A.

How requests get prioritized: upvotes are the loudest signal. No promises, but I read everything.

Welcome! I'm glad you're here.