ChessBench

[PREVIEW]

A New Chess Benchmark for Language Models

Support ChessBench

Independently benchmarking a language model takes hundreds of games and, for newer frontier models, costs hundreds — sometimes thousands — of dollars in API fees. Your support helps test more models, faster and more thoroughly.

GitHub Sponsors

One-time or monthly support with tiers and a public sponsor badge.

Become a sponsor on GitHub →

Stripe

Pay any amount, no account required.

Make a one-time contribution, or give monthly: $5 · $10 · $20 · $50

Community

Tell us which models or matchups you'd like to see, ask questions, share findings. Community input shapes which models we prioritize for benchmarking.

Join the discussion on GitHub →

Or follow along for new results and announcements: LinkedIn · Bluesky · Twitter


ChessBench is currently operated by an individual; contributions are not tax-deductible.

Funds go directly to API costs and project time.

Thanks for considering it.

Benjamin Brumfield