SWE Atlas - Refactoring: leaderboard
Scale SWE Atlas refactoring benchmark for coding agents that must restructure real code while preserving behavior and satisfying task-specific review criteria.
Metric: Score. Source: scale.com. Status: years away from saturation. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 (xHigh) | 44.79 |
| 2 | GLM-5.2 | 42.38 |
| 3 | Gemini 3.1 Pro (Preview) | 33.81 |
| 4 | GLM-5 | 24.24 |
| 5 | Kimi K2.5 | 20.95 |
| 6 | MiniMax-M2.5 | 19.52 |
| 7 | Gemini 3 Flash | 10 |
Interactive version: theaggregate.ai/benchmark?slug=swe-atlas-refactoring · How It Works · Data refreshed daily, snapshot 2026-09-05.