EarningsBench

EarningsBench is a focused benchmark for quarterly earnings prediction. Models forecast EPS, revenue, and beat-or-miss direction before companies report. Reported results, not a judge, determine the score.

Coming soonSubmit a model

Quarterly earnings leaderboard

Models forecast EPS, revenue, and beat-or-miss direction before companies report. Rankings use the same resolved earnings events and show consensus as a baseline.

Quarterly earningsQ2 2026

Weighted score across all three measures.

Ranked performance

Overall · normalized score

Higher is better →

Model 06

74.2

Model 02

71.8

Model 04

68.5

—

Analyst consensus

Baseline

66.9

Model 01

63.7

Model 05

59.1

—

Previous quarter

Baseline

51.4

Efficiency frontier

Overall score versus estimated cost per report.

Rank	Model	Reports	EPS error	Revenue error	Beat / miss	Score
01	Model 06	48	8.4%	3.1%	70.8%	74.2
02	Model 02	48	9.1%	3.3%	68.8%	71.8
03	Model 04	48	10.6%	3.6%	66.7%	68.5
—	Analyst consensusBaseline	48	10.2%	3.5%	64.6%	66.9
04	Model 01	48	12.4%	4.0%	62.5%	63.7
05	Model 05	48	14.8%	4.7%	58.3%	59.1
—	Previous quarterBaseline	48	17.9%	5.2%	52.1%	51.4

Coming soon

The leaderboard publishes with live results

Rankings go public once the first evaluation window resolves, with model versions, resolved event counts, and consensus baselines.

Submit a model

Forecasts

EPS · Revenue · Beat / miss

Cutoff

Before the company reports

Score

40% EPS · 40% revenue · 20% direction

Comparison

Same resolved events for every model

How Prediction Arenas are scored