AA Index · data in top since 09.09.2026vibecoding.tech_

AI model rankings

Artificial Analysis Coding Index · artificialanalysis.ai · 14 models · data in top since 09.09.2026

AI agents Models MCP

Model

Claude Fable 5.1max

81.6

—

$50

GPT-5.6 Solxhigh

78.3

—

$30

Claude Opus 5max

—

$25

gpt-6-astrahigh

77.1

—

Grok 4.6high

76.8

—

GPT-5.6 Terramax

76.7

—

$15

Claude Fable 5max

76.5

—

$50

Muse Spark 1.3xhigh

76.5

—

Gemini 3.8 Flashhigh

76.3

—

$3.75

Kimi K3max

76.2

—

$15

Gemini 3.7 Flashhigh

76.1

—

GLM-5.3max

74.8

—

$4.4

Qwen3.8-Flash-Next

73.1

—

Grok 4.5high

72.4

—

020406080100

result - metric from the active source · price per task/case - only if published by the same source · $/1M - separate output-token rate · the best config is selected again for each source

How to read the table

four tabs - four rankings

Artificial Analysis, DeepSWE, SWE-Rebench, and Aider are not combined: each tab shows up to 14 models.

configuration belongs to the source

one model can win with different effort/config. The bar, badge, and row order change with the source.

price is not combined

price/task is shown only when the active source publishes it; $/1M out - separate pricing metadata, not part of the benchmark score.

Vibecoding AI Coding Index - an internal composite of four signals, not a fifth source or a tab in this ranking.

Editorial metadata (curated metadata) add pricing, provider, context, and release date. These fields enrich the row but do not move it in the source ranking.

Methodologywhere the figures come from and how to read the four sourcesRuBenchour benchmark on Russian technical specifications: pass@1 with open trajectoriesAll rankingsmodels, AI agents, MCP