About Funds Benchmarks Insights Contact
Invest

Intelligent Earnings Benchmark

AI predicts the direction of earnings.

We test frontier AI models on the hardest question in equity research: How will forward earnings expectations change before a stock reprices it?

Results

Model leaderboard

Methodology

How the benchmark works.

The Intelligent Earnings Benchmark tests whether frontier AI models can predict where the market's forward earnings expectations are too high or too low. Each model receives identical company data and is asked to predict the direction and magnitude of forward estimate revisions over the next 60 days.

Data Provided

8 quarters of income statements and balance sheets, 12 quarters of earnings surprise history, the most recent earnings call transcript, current consensus estimates, and 8 quarters of historical consensus revision data.

Tools Available

All models share access to historical financial data, recent company commentary, FRED economic data, and a calculation tool. Every model receives the same directions and base-rate details.

What Models Predict

Revenue and EPS revision direction (up, down, flat), magnitude (small, medium, large), an implied revision percentage, confidence score, key signals, and overall rationale.

Scoring

Predictions are scored against realized consensus changes measured 60 days after the prediction deadline. Directional accuracy measures the right call; magnitude accuracy measures sizing.

Universe

US-listed common equities with market caps above $10B, excluding ADRs, SPACs, and ETFs. Frozen at each quarterly deadline and re-screened quarterly.

Auditability

Every run is logged for the entire universe with the prompt, data packets, and predictions. Model version strings and tool-call logs are preserved and re-scorable.

Subscribe

The future of investing is intelligent.

Follow our work — research updates and new findings on how advanced AI invests, straight to your inbox.