Matrix/Feature/SWE-bench entry
SWE-bench entry: which coding agents have it?
A submission on the official SWE-bench leaderboard (swebench.com data file). The product, not just its model, must be named.
3 Yes1 Partial0 Unknown
| Agent | Value | Notes | Source |
|---|---|---|---|
| Factory DroidsFactory | 37.0% Verified (2024) | Factory Code Droid: 37.0% SWE-bench Verified, 31.33% Lite, 19.27% Test (all submitted 2024-06-17). An old entry. | Source |
| JulesGoogle | 52.2% Verified (2024) | Google Jules + Gemini 2.0 Flash (experimental) resolved 52.2% of SWE-bench Verified, submitted 2024-12-12. An old entry. | Source |
| OpenHands CloudAll Hands AI | 71.8% Verified (2025) | OpenHands entries on SWE-bench Verified include 71.8% with GPT-5 (2025-08-07) and 70.4% with Claude 4 Sonnet (2025-05-24). | Source |
| Codex cloudOpenAI | Model only: 72.8% Verified | Not the Codex cloud product: the GPT-5.2 Codex model scores 72.8% on SWE-bench Verified (2026-02-19) and 66.3% Multilingual inside the mini-SWE-agent harness. | Source |
| AmpAmp (Sourcegraph spin-out) | None listed | No Amp entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| Base44Base44 (Wix) | None listed | No Base44 entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| Bolt.newStackBlitz | None listed | No Bolt entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| Claude Code on the webAnthropic | None listed | No Claude Code entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| Cursor Cloud AgentsAnysphere | None listed | No Cursor entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| DevinCognition | None listed | No Devin entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| Firebase StudioGoogle | None listed | No Firebase Studio entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| GitHub Copilot cloud agentGitHub | None listed | No Copilot entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| KiroAWS | None listed | No Kiro entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| LovableLovable | None listed | No Lovable entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| Replit AgentReplit | None listed | No Replit entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
| v0Vercel | None listed | No v0 entry on the official SWE-bench leaderboards (Verified, Lite, Test, Multilingual, Multimodal) as of the checked date. | Source |
Other features
Weekly, free
Which coding agent changed this week
One short email when coding agents and app builders change prices, limits, MCP support or benchmark results. Every item links its source.