AIEvals

Benchmarks

Artificial Analysis Intelligence Index

An index over a fixed suite of reasoning, coding, mathematics and agentic evaluations, reported on its own scale rather than as a percentage.

CompositeArtificial Analysis492 models run by Artificial AnalysisUpdated 8 October 2026

How it is run

Artificial Analysis runs each model on its own harness at several reasoning-effort settings and publishes the highest-effort result it measured, named in the table. Version 4.3.2 of the index adds AutomationBench-AA, moves Terminal-Bench to 4.0 and drops τ³-Banking; cost per task is measured on the same harness.

What it cannot tell you

A composite over one publisher's suite, so it says nothing about Vals AI's suite and cannot be averaged with the Vals Index: their task mix, weights and harnesses differ. The number also depends on which effort setting was published, and a single component can move the index.

Leader
Claude Opus 5.558 · among the models on this site
Top-ten spread
10Leader minus tenth place, or last place in smaller fields.
Median
42of the 42 models on this site
No published result
4of 46 models on this site

No top list to rank

Artificial Analysis scores 492 models on this benchmark but publishes no ordered list for them, so this site ranks only the 42 models it holds a number for and claims nothing beyond that.

Models tracked on this site

All 42 models this site has a published number for. Ranks are among these models only, within the 492 models measured by Artificial Analysis.

# hereModelIndex
1Claude Opus 5.558
2Claude Sonnet 5.556
3Claude Fable 5.153
3Gemini 4 Argon53
3GPT-6 Astra53
6GPT-6.1 Sol52
7Claude Opus 551
8Claude Fable 550
9GPT-6 Sol48
9Muse Spark 1.3 Max48
11GPT-5.6 Sol47
12Grok 4.746
12MiMo V2.6 Pro46
14GLM 5.345
14Qwen3.8 Max (0902)45
16Grok 4.644
16Kimi K344
16Step 5 Preview44
19Claude Haiku 5.543
20Claude Opus 4.842
20GLM 5.3 Flash42
20GPT-5.6 Terra42
23Claude Opus 4.741
23Gemini 3.8 Flash41
25Muse Spark 1.240
25Qwen3.8 2.4T A95B40
25Qwen3.8-Flash-Next40
25Qwen3.8 Max40
29DeepSeek V4.1 Flash39
29Gemini 3.7 Flash39
29GPT-5.439
29Grok 4.539
33Claude Sonnet 538
33GPT-5.538
33MiMo-V2.6-Flash38
36GPT-5.6 Luna37
36GPT-6 Luna37
38DeepSeek V4 Pro36
39Gemini 3.1 Pro30
40MiniMax M329
41Inkling25
42Mistral Medium 3.514

Source: Artificial Analysis · updated 8 October 2026

No published result tracked

4 of the 46 models on this site have no published number on this benchmark, so their cells in the home table stay blank rather than reading as zero.

DeepSeek V4.1 Flash 1017GLM-LegalBeamHunyuan Hy4 Preview

Read this with

Benchmarks in the same category, which usually disagree with this one on purpose.

  • Vals Multimodal IndexSame category. A weighted composite of the multimodal tasks in the independent suite, intended as a single number for image-and-document work.
  • Vals IndexSame category. A GDP-weighted composite of Vals AI’s agentic finance, coding, legal and tax evaluations, intended to represent potential economic impact.

Sources

Every page used for this benchmark, read on 17 October 2026.