Benchmarks › AA Decision Index

AA Decision Index

Choose answers in classification, tool calling, knowledge and reasoning tests
65.2top score · D1 Ultra
15models ranked
6open-weights models
Weighted composite indexhigher is better
Oct 21, 2026last updated

How the test works

1

What it is

Choose answers in classification, tool calling, knowledge and reasoning tests

  • Kinds of tasksClassification, tool calling, knowledge, reasoning
2

Components

5 parts

  • WeightsAA-Omniscience-Decision accuracy 12.5%, AA-Omniscience-Decision non-hallucination rate 12.5%, HLE-MultiChoice 25%, When2Call 25%, WildGuardTest 25%
  • How runs combineWeighted mean of four evaluations; Omniscience contributes two 12.5% terms. All questions in a task are answered in one call.
3

Score

Weighted composite index

  • Better isHigher

AA Decision Index v1.0 leaderboardtop 10 of 15

Full leaderboard with filters →
ModelAA Decision Index v1.0Per taskPriceSpeedSupportsContext
D1 UltraLiquid AI
65.2
—$0.042 / ——33K
TJev 1.13TypeSafe
57.5
—$0.042 / ——64K
TJev 1.14TypeSafe
57.1
—$0.042 / ——64K
Decider v1.2PerplexityOpen
56.2
—$0.02 / ——262K
FGLiDEFastino
53.9unstated
—$0.15 / ——40K
53.8
—$0.042 / ——33K
Decider v1.1 27BPerplexityOpen
53.1
——8K
52.7
—$0.02 / ——66K
52.3
—$0.04 / ———
D1Liquid AI
50.3
—$0.02 / ——66K

AA Decision Index questions

What is AA Decision Index?›

Choose answers in classification, tool calling, knowledge and reasoning tests. Created by Artificial Analysis.

Which model is best on AA Decision Index?›

D1 Ultra leads AA Decision Index v1.0 with 65.2, ahead of Jev 1.13 (57.5) and Jev 1.14 (57.1).

What is the best open-weights model on AA Decision Index?›

Decider v1.2 with 56.2, then Decider v1.1 27B (53.1).

How is AA Decision Index scored?›

Weighted composite index; four evaluations, 25% weight each.

Which versions of AA Decision Index are there?›

AA Decision Index v1.0 (15 models), AA Decision · HLE (MCQ) (15 models), AA Decision · Omniscience (15 models), AA Decision · Non-hallucination (15 models), AA Decision · When2Call (15 models), AA Decision · WildGuardTest (15 models), AA Decision cost per task (15 models), AA Decision time per task (15 models).