AI Agent Benchmarks & Leaderboards
Data-driven performance comparisons across AI agents. We track SWE-bench coding scores, response latency, token pricing efficiency, browser automation success rates, and more. Updated weekly.
SWE-bench Coding Leaderboard
SWE-bench evaluates AI agents on real GitHub issues. Higher scores mean the agent successfully resolves more real-world coding tasks.
Rank
Agent
SWE-bench Score
Rating
Pricing
#1
DevinThe first AI software engineer by Cognition AI
74%
4.9
$500-5000/mo
#2
Claude CodeAnthropic's official agentic coding tool in your terminal
49%
4.8
$20-200/mo
#3
ClaudeAnthropic's AI assistant for thoughtful conversation
49%
4.8
$0-200/mo
#4
AiderAI pair programming in your terminal
46.1%
4.7
Free (API costs apply)
#5
CursorThe AI-first code editor built on VS Code
44.5%
4.7
$0-20/mo
#6
CursorThe AI-first code editor built on VS Code
44.5%
4.7
$0-20/mo
#7
ClineAutonomous coding agent for VS Code
43%
4.8
Free (API costs apply)
#8
SWE-AgentOpen-source autonomous software engineering agent
42.7%
4.6
Free (API costs apply)
#9
WindsurfCodeium's AI-powered IDE with agentic capabilities
41%
4.6
$0-15/mo
#10
ChatGPTOpenAI's conversational AI platform
40%
4.6
$0-200/mo
Benchmark Categories
Most Cost-Efficient AI Agents
Token pricing comparison (input cost per 1M tokens). Lower costs mean more affordable API usage.