About Model Madness

Model Madness is an experiment by Alephic to see how well AI language models can predict the NCAA March Madness tournament.

The Experiment

We asked 45 different AI models from providers like Anthropic, OpenAI, Google, Meta, Mistral, and others to fill out March Madness brackets. Each model runs two strategies:

• One-shot — The model fills out the entire bracket at once, predicting all 63 games in a single pass.
• Round-by-round — The model predicts one round at a time, receiving actual results before predicting the next round.

Scoring

Points increase with each round to reward correctly predicting later, harder games:

• Round of 64: 10 points per correct pick
• Round of 32: 20 points
• Sweet 16: 40 points
• Elite 8: 80 points
• Final Four: 160 points
• Championship: 320 points

About Alephic

Alephic is an AI consultancy that helps companies build with large language models. Model Madness is one of our explorations into how different models reason about uncertain, real-world predictions.