Dividend Intelligence Benchmark

HeyDividend AI Benchmark

How does HeyDividend AI stack up against GPT-4o, Claude, Gemini, Grok-4, and other leading models on real-world dividend intelligence tasks? We tested 11 AI models on real dividend accuracy.

UPDATED Live
MODELS 11
PILLARS 4
SOURCE HeyDividend AI

Headline

HeyDividend AI Composite Score
62.4
Ranked #1 of 11 models
Nearest Competitor
46.7
Perplexity Sonar
Lead Margin
+15.7
points ahead

HeyDividend AI's Structural Moat

HeyDividend AI achieves 100% ex-date accuracy through a database-first architecture — sourcing dividend dates directly from authoritative financial databases, not LLM generation. No competitor can close this gap through training alone, as even the best LLMs achieve ≤55% accuracy on ex-date prediction.

Scoring Formula

WEIGHT: 40%
Dividend Cut F1
51.0%
Precision × recall on predicting which companies will cut dividends. HeyDividend AI uses ML models trained on financial fundamentals.
WEIGHT: 30%
Yield MAPE
40.0%
Mean Absolute Percentage Error on dividend yield predictions. Lower is better — converted to accuracy via (1 − MAPE/100).
WEIGHT: 30%
Ex-Date Accuracy
100%
Percentage of ex-dividend dates reported correctly. HeyDividend AI's DB-first approach gives it a structural 100% on this pillar.
composite = 0.40 × cut_F1 + 0.30 × (1 − MAPE/100) + 0.30 × ex_date_accuracy

Models Compared

HeyDividend AI (winner)
Strong (≥40)
Moderate (≥25)
Weak (<25)

Composite Score Leaderboard

Composite Score
Weighted combination of Cut F1, Yield MAPE accuracy, and Ex-Date accuracy
OVERALL

Per-Pillar Breakdown

Dividend Cut F1
Precision × recall on predicting which companies will cut dividends
WEIGHT: 40%
Yield Accuracy (1−MAPE/100)
Mean Absolute Percentage Error on yield predictions — lower MAPE = higher accuracy
WEIGHT: 30%
Ex-Date Accuracy
Percentage of ex-dividend dates reported correctly
WEIGHT: 30%

Full Scorecard

Coverage & Data

SYMBOLS IN DB
20,624
dividend stocks
PAYMENTS ANALYZED
783,022
dividend records
AVG HISTORY
38
payments per symbol
EX-DATE ACCURACY
100%
DB-first precision

Methodology

How We Score

Scoring Formula

0.40 × cut_F1 + 0.30 × (1 − MAPE/100) + 0.30 × ex_date_accuracy. Scale: 0 – 100.

HeyDividend AI Scoring

Backtested on 30 years of production dividend data from our unified dividends database.

Competitor Scoring

Estimated from FinanceBench 2024, FLARE-FinQA, published model papers, and structural analysis.

Update Cadence

Re-run monthly across all three pillars using the latest available model versions for fair, current comparisons.

Frequently Asked Questions

Which AI is most accurate for dividend intelligence?
HeyDividend AI leads with a composite score of 62.4, outperforming all 10 competitors tested. The nearest competitor, Perplexity Sonar, scores 46.7 — a gap of 15.7 points. Other models tested include Grok-4 (39.2), BloombergGPT (35.9), Gemini 2.5 Pro (28.7), Claude Sonnet (28.1), GPT-4o (26.3), DeepSeek-R1 (14.2), and FinGPT (16.3).
How is the AI dividend benchmark scored?
The composite score uses three weighted pillars: Dividend Cut F1 (40% weight), Yield MAPE (30% weight), and Ex-Date Accuracy (30% weight). Formula: composite = 0.40 × cut_F1 + 0.30 × (1 − MAPE/100) + 0.30 × ex_date_accuracy.
Why does HeyDividend AI score higher than GPT-4o and Claude on dividends?
HeyDividend AI achieves 100% ex-date accuracy through a database-first architecture rather than generating dates with an LLM. Even the best general-purpose LLMs achieve ≤55% accuracy on ex-date prediction. Combined with ML models trained on financial fundamentals for cut prediction, this creates a structural advantage.
How many AI models are compared in the benchmark?
11 AI models: HeyDividend AI, Perplexity Sonar, Grok-4, Perplexity Finance, BloombergGPT, Gemini 2.5 Pro, FinGPT, Claude Sonnet, GPT-4o, DeepSeek-R1, and a Naive Baseline. Updated monthly.
What is a Dividend Cut F1 score?
It's the harmonic mean of precision (how many predicted cuts actually happened) and recall (how many actual cuts were predicted). Most general-purpose LLMs score below 40% on this specialized financial task.
How often is the benchmark updated?
Monthly, re-running the full evaluation across all three pillars using the latest available model versions.

The engine behind the score

Put the #1-ranked dividend AI to work

The benchmark is the proof. HeyDividend AI runs the same models against your real portfolio — with cut-risk alerts, NAV verdicts, and tax-aware research on top.

See pricing Sign up free →