SWE-bench Live Lite (SWE-agent): leaderboard
SWE-bench-Live (Microsoft, 2025) turns recent GitHub issues into runnable tasks with its automated RepoLaunch pipeline, and a patch counts only when the repository's tests pass. This board is the frozen Lite split (300 Python tasks, 50 a month from issues opened between October 2024 and March 2025) under the SWE-agent harness, scored as the percentage of tasks resolved; higher is better.
Metric: Resolved rate (%). Source: swe-bench-live.github.io. Saturation forecast: Around March 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 36 | #244 |
| 2 | GPT-5 (Medium) | 24.33 | #163 (GPT-5) |
| 3 | Claude 3.7 Sonnet (20250219) | 17.67 | #399 |
| 4 | GPT-4.1 | 16.33 | #492 |
| 5 | DeepSeek V3 (0324) | 15.33 | #704 |
| 6 | GPT-4o (2024-11-20) | 10 | #864 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=swe-bench-live-lite-swe-agent · How It Works · Data refreshed daily, snapshot 2026-11-12.