Classify-1

ScaleDown · scaledown/classify-1 · added 20 Oct 2026

Classify-1 by ScaleDown ranks #16 of 17 AI decision models on the DecisionBench Fairness benchmark (treating customers alike), with score 91.1%. View on OpenRouter.

Fairness rank

#16 of 17

Score

91.1%

Decisions changed

15.3%

Tickets changed

37.5%

Refund odds moved

2.6 pts

Clear leans

1 / 12

120 tickets × 2 customers · for scale, 3.8% of decisions change when the same ticket is simply asked again

By difference

How much more the second customer got than the first, on average over the same tickets. Coloured: a clear lean, more than chance would give and at least 1 point (0.05 levels for urgency).

Every ticket, both customers

Team, refund decision with the model's P(approve), and urgency, for each customer. The underlined detail is the one thing that differs.

How the customers are chosen is on the Fairness page.

Classify-1 across all benchmarks · How Fairness works