---
title: "JetBrains Kotlin Benchmark: Claude Code tops 85.71% on 105 tasks"
slug: "jetbrains-kotlin-benchmark-claude-code-tops-8571-on-105-tasks"
published: "2026-08-20"
beat: "Research"
tags: ["Research"]
creator: "Agentry Newsroom"
editor: "Susanne Sperling, Editor — Human in the Loop"
tools: ["Claude (Anthropic)", "Perplexity Sonar"]
creativeWorkStatus: "verified"
dateReviewed: "2026-08-20"
aiActArticle50: "compliant"
humanView: "https://agentry.news/research/jetbrains-kotlin-benchmark-claude-code-tops-8571-on-105-tasks"
agentView: "https://agentry.news/agent/jetbrains-kotlin-benchmark-claude-code-tops-8571-on-105-tasks"
---# JetBrains Kotlin Benchmark: Claude Code tops 85.71% on 105 tasks

> JetBrains published a real-world Kotlin benchmark on July 23, 2026, measuring AI coding agent performance across 105 engineering tasks from open-source repositories. Claude Code with Opus 4.7 xhigh le

*Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. [AI policy](/ai-policy).*

JetBrains published the Kotlin Benchmark for AI coding agents on [July 23, 2026](https://blog.jetbrains.com/zh-hans/kotlin/2026/07/introducing-the-kotlin-benchmark-evaluate-ai-coding-agents-on-real-world-kotlin-tasks/), introducing a standardized evaluation framework designed to measure how AI agents perform on authentic Kotlin development work.

## Benchmark structure and scope

The benchmark comprises **105 engineering tasks** drawn directly from open-source Kotlin repositories, ensuring that evaluation criteria reflect real-world development scenarios rather than synthetic or toy problems. By grounding the assessment in actual codebases, JetBrains created a measurement tool applicable to production environments where coding agents are increasingly deployed.

## Leaderboard results

**Claude Code paired with Opus 4.7 xhigh** claimed the top position on the leaderboard, successfully resolving **90 of 105 tasks** for a documented success rate of **85.71%** [JetBrains](https://blog.jetbrains.com/zh-hans/kotlin/2026/07/introducing-the-kotlin-benchmark-evaluate-ai-coding-agents-on-real-world-kotlin-tasks/).

The closest competitors followed substantially behind: JetBrains Junie paired with Opus 4.7 max achieved 81.9%, as did Codex with GPT 5.5 xhigh, both holding equal runner-up positions [JetBrains](https://agentry.news/launches/jetbrains-kotlin-benchmark-claude-code-leads-at-8571). The 3.81 percentage-point gap between first and second place underscores measurable performance differentiation across agent-model combinations in a language-specific coding context.

## Market significance

The release arrives amid accelerating adoption of AI coding agents across enterprise development teams. Benchmarks that isolate performance by programming language provide enterprises and development tool vendors with concrete comparison data for agent selection and integration decisions. JetBrains' decision to publish this evaluation against its own agent offering—Junie—alongside competitors reflects a shift toward transparent, quantifiable agent performance metrics.

The Kotlin-specific focus addresses a segment of the developer population that relies on the JVM ecosystem. As organizations scale agent deployment, language-targeted benchmarks reduce abstraction and allow teams to predict agent behavior within their actual tech stacks.

## Next steps for agent vendors

Published benchmarks establish de facto testing grounds for agent refinement. Vendors whose agents underperformed have concrete targets for improvement, while leading agents gain credibility through third-party validation. JetBrains' framework may prompt similar benchmarking efforts from other language maintainers and IDE vendors seeking to quantify agent capability in their ecosystems.