Benchmarks › AA Intelligence Index

AA Intelligence Index

By Artificial Analysis ↗
Combine results from ten knowledge, reasoning, coding and agent evaluations
57.6top score · Claude Opus 5.5
334models ranked
223open-weights models
Weighted composite indexhigher is better
Oct 20, 2026last updated

How the test works

1

What it is

Combine results from ten knowledge, reasoning, coding and agent evaluations

  • Kinds of tasksComposite evaluation index
2

Components

11 parts

  • WeightsAA-Briefcase v1.1 15%, GDPval-AA v2.1 10%, AutomationBench-AA 5%, Terminal-Bench 4.0 10%, SciCode 10%, AA-Omniscience accuracy 10%, AA-Omniscience non-hallucination rate 5%, GDP.pdf 10%, AA-LCR v1.1 5%, Humanity's Last Exam 10%, CritPt 10%
  • How runs combineWeighted mean of eleven scored terms across ten evaluations
3

Score

Weighted composite index

  • Better isHigher
Sources: artificialanalysis.ai

AA Intelligence Index 4.3.2 leaderboardtop 10 of 334

Full leaderboard with filters →
ModelAA Intelligence Index 4.3.2Per taskPriceSpeedSupportsContext
Claude Opus 5.5Anthropic
57.6max
$5.98$4 / $20961M
Claude Sonnet 5.5Anthropic
56max
$5.46$2 / $101371M
Muse Wonder FullMeta
55.8
——1M
Claude Fable 5.1Anthropic
53.4max
$7.63$10 / $50691M
GPT-6 AstraOpenAI
52.7max
$3.26$10 / $50691.05M
Gemini 4 ArgonGoogle
52.6high
$1.99$2 / $10841M
Muse WonderMeta
52.4max
$4.52$4 / $201611M
GPT-6.1 SolOpenAI
51.8max
$0.72$2 / $10521.05M
Claude Opus 5Anthropic
50.8max
$5.86$5 / $25551M
Claude Fable 5Anthropic
49.6max
$8.75$10 / $50671M

AA Intelligence Index questions

What is AA Intelligence Index?›

Combine results from ten knowledge, reasoning, coding and agent evaluations. Created by Artificial Analysis.

Which model is best on AA Intelligence Index?›

Claude Opus 5.5 leads AA Intelligence Index 4.3.2 with 57.6, ahead of Claude Sonnet 5.5 (56) and Muse Wonder Full (55.8).

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

MiMo v2.6 Pro with 46.3, then DeepSeek V4.1 Flash 1017 (45.1).

How is AA Intelligence Index scored?›

Weighted composite index; agents 30%, coding 20%, scientific reasoning 20%, general 30%.