Agent Leaderboard

Coding agents ranked by SWE-bench Verified resolve rate.

Source: SWE-bench

#	Agent	Model	Resolve %	$/instance	Type
1	Claude CodeAnthropic	Claude Sonnet 4	72.5%	$4.38	Closed
2	DevinCognition	Devin Agent	71.6%	$5.21	Closed
3	OpenHandsAll Hands AI	Claude Sonnet 4	69.1%	$3.87	OSS
4	SWE-agentPrinceton NLP	Claude Opus 4	64.8%	$2.15	OSS
5	AiderAider AI	Claude Opus 4	62.3%	$1.92	OSS
6	Cursor AgentCursor	Custom	60.1%	-	Closed
7	Copilot WorkspaceGitHub	GPT-4.1	55.4%	-	Closed
8	ClineCline	Claude Sonnet 4	52.8%	$3.14	OSS
9	WindsurfCodeium	Custom	48.2%	-	Closed
10	AutoCodeRoverNUS	GPT-4o	43.6%	$0.65	OSS

Scores from SWE-bench Verified. Cost is average per instance in USD.