Skip to content

BenchmarksArtificial Analysis

Artificial Analysis Coding Agent Index v2.1: Claude Fable 5.2 and Opus 5.5 lead at 58, cost per task spans 29×

Coding Agent Index at a glance

Artificial Analysis released the Coding Agent Index v2.1 on November 13, 2026, rebuilding its coding agent benchmark on FrontierCode 2.0 and Terminal-Bench 4.1. As of November 14, 2026, Claude Fable 5.2 (58.5) and Claude Opus 5.5 (58.2) lead the 29 configurations, while cost per task spans $1.51 to $44.22.

Current standing

As of

Rank
#1of 47 models

Claude Fable 5.2 (Max) · 64.6 on the Artificial Analysis Intelligence Index v5.0.1 (opens in a new tab)

Current leader
Still the top model in this snapshot.

Computed from our daily snapshot; it changes as new models are measured. Artificial Analysis Coding Agent Index v2.1: Claude Fable 5.2 and Opus 5.5 lead at 58, cost per task spans 29×

Ask your AI about this story

Use this story in your business: Artificial Analysis Coding Agent Index v2.1: Claude Fable 5.2 and Opus 5.5 lead at 58, cost per task spans 29× (Artificial Analysis). Ask your AI to explain what happened, how it could affect your business and how to put it to work. It also weighs price, speed, intelligence and the lab behind it. The link to the story is included.

What changed with Coding Agent Index

Artificial Analysis released version 2.1 of its Coding Agent Index on November 13, 2026. The Index is the equal-weight average of three benchmarks across 322 tasks: FrontierCode-AA 2.0 (132 software engineering tasks, Artificial Analysis's implementation of Cognition's FrontierCode 2.0, graded on whether a maintainer would merge the change), Terminal-Bench 4.1 (66 agentic terminal tasks) and SWE-Atlas-QnA-AA (124 repository question-and-answer tasks). Each benchmark score averages three attempts per task. FrontierCode 2.0 draws tasks from 55% more repositories than version 1.1, and its reference solutions modify 67% more files and 89% more lines of code. Terminal-Bench 4.1 keeps the 66 tasks of 4.0 and updates many verifiers, running agent code as an untrusted user during grading to resist reward hacking.

As of November 14, 2026, the Index ranks 29 configurations. Claude Code running Claude Fable 5.2 at xhigh effort leads at 58.5, ahead of a Devin Fusion CLI combination (Claude Opus 5.5 Max with GPT-6 Luna High) at 58.4, Claude Fable 5.2 at high (58.2) and Claude Opus 5.5 at max (58.2), both in Claude Code. Fable 5.2 also leads FrontierCode-AA 2.0 at 49.2. OpenAI's GPT-6.1 Sol in Codex scores 52.2. The cost spread is the story: among configurations scoring 48 or higher, average cost per task runs from $1.51 (Codex with GPT-6.1 Sol at max) to $44.22 (Claude Code with Fable 5.2 at max), about 29 times more. Fable 5.2 in Claude Code costs $2.89 per task at medium effort, $7.51 at high and $15.47 at xhigh.

Coding Agent Index in numbers

Values are copied from the linked source on the date shown. They are not IA Compa measurements.

Coding Agent Index v2.1 score (Claude Code, xhigh) · Claude Fable 5.2as of · Source for Coding Agent Index v2.1 score (Claude Code, xhigh): artificialanalysis.ai (opens in a new tab)
58.5
Coding Agent Index v2.1 score (Claude Code, max) · Claude Opus 5.5as of · Source for Coding Agent Index v2.1 score (Claude Code, max): artificialanalysis.ai (opens in a new tab)
58.2
Coding Agent Index v2.1 score (Codex, max) · GPT-6.1 Solas of · Source for Coding Agent Index v2.1 score (Codex, max): artificialanalysis.ai (opens in a new tab)
52.2
FrontierCode-AA 2.0 score (Claude Code, xhigh) · Claude Fable 5.2as of · Source for FrontierCode-AA 2.0 score (Claude Code, xhigh): artificialanalysis.ai (opens in a new tab)
49.2
Average cost per task (Claude Code, max) · Claude Fable 5.2as of · Source for Average cost per task (Claude Code, max): artificialanalysis.ai (opens in a new tab)
$44.22
Average cost per task (Codex, max) · GPT-6.1 Solas of · Source for Average cost per task (Codex, max): artificialanalysis.ai (opens in a new tab)
$1.51

Where to compare Coding Agent Index

Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.

Why Coding Agent Index matters for businesses

Why it matters: a version bump like this resets every comparison a business made last quarter — scores under v2.1 are not comparable with v2.0, and the ranking of the harnesses you can actually buy has changed. The number that should drive the purchase is cost per completed task: configurations within about six points of the lead differ in price by a factor of roughly 29, and the leading model at a mid effort setting (Fable 5.2 at high, $7.51 per task) outscores its own max setting while costing about a sixth as much. If you buy coding agents for a team, or quote agent-built work to clients, price the job from cost-per-task data and re-check it after every Index version change.

Put it to work: AI services

Get AI News by email

Verified AI stories, each with its official source, on Monday and Thursday mornings. Add breaking news to hear about the biggest ones as soon as we publish them.

Every story links its official source

Free, in English or Spanish

Unsubscribe with one click, any time

What to send

Sources for this story

  1. Official
  2. Official
  3. FrontierCode 2.0 (opens in a new tab)

    Cognition · cognition.com

    Official
  4. Terminal-Bench 4.1 (opens in a new tab)

    Terminal-Bench (Laude Institute) · tbench.ai

    Official

Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.

  • #coding-agent-index
  • #frontiercode
  • #terminal-bench
  • #coding-agents
  • #claude-fable-5-2
  • #cost-per-task
  • #artificial-analysis
  • #benchmarks

Want this working in your business?

We turn new AI capabilities into agents, voice receptionists, and automations you can actually use.