Benchmark
PolymarketBench
Track AI models making forecasts through simulated prediction-market trades.
Explore the original evidence
Explore the original evaluation, its methods, and reported findings.
Open evalCoverage
Seven models in a 30-day paper-money run.
Methodology
The source specifies $1,000 per model, $20 per market per day, with bets and sessions published after 48 hours.