AI benchmarks, answered by you
Try the questions behind the scores.
Pick a subject, work through a short session, and review the answers. No account or timer. Your place is saved in this browser.
Loading the collection...
How practice works
Multiple-choice questions use the published answer key. Math questions in GSM8K accept equivalent numbers, including fractions. For written and code answers, reveal the reference and assess your own work. Self-assessments are kept separate from automatic scores.
Counts describe the questions included here. Open a benchmark for its split and source details, or browse all sources and scoring notes.