Drex Max

Nace.AI · nace-ai/drex-max · added 22 Oct 2026

Drex Max by Nace.AI ranks #2 of 18 AI decision models overall on DecisionBench: #11 of 18 on Sudoku, #1 of 18 on Rulebook, #3 of 18 on Consistency, #2 of 18 on Odds, #1 of 18 on Fairness. View on OpenRouter.

Overall: #2 of 18
Overall score: 84 / 100
Sudoku: #11 of 18
Rulebook: #1 of 18
Consistency: #3 of 18
Odds: #2 of 18
Fairness: #1 of 18

Logical reasoningSudoku#11

Accuracy: 78.9%
Solved: 20/20
Mistakes / puzzle: 13.4

Full Sudoku results →Following written rulesRulebook#1

Accuracy: 99.4%
Confidence gap: 0.005
Cost / 1k tickets: $0.0643

Full Rulebook results →Not getting thrown offConsistency#3

Score: 99.5%
Same if re-asked: 100.0%
Obeyed planted text: 0.0%

Full Consistency results →Weighing riskOdds#2

Score: 55.5%
Value captured: 76.0%
Odds within 2×: 35.0%

Full Odds results →Treating customers alikeFairness#1

Score: 99.8%
Decisions changed: 0.3%
Clear leans: 0 / 12

Full Fairness results →

On each benchmark the best model scores 100 and the worst 0. The overall score averages all 5, so every benchmark counts the same. Rerun 1,000 times on redrawn test sets, it finishes #2 in 97% of them, in tier 2 on its own. How ranking works.