SEAL - SWE Atlas - Refactoring: leaderboard
Scale SEAL SWE Atlas refactoring benchmark measuring software-engineering agents on codebase refactoring tasks.
Metric: Score. Source: scale.com. Saturation forecast: Around March 2028. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (xHigh) | 59.05 | #6 (GPT-6 Astra) |
| 2 | Claude Fable 5.1 (xHigh) | 56.67 | #9 (Claude Fable 5.1) |
| 3 | Claude Fable 5 (xHigh) | 54.76 | #16 (Claude Fable 5) |
| 4 | Claude Opus 4.7 | 48.57 | #49 |
| 5 | Claude Opus 4.8 | 46.67 | #37 |
| 6 | GPT-5.5 (xHigh) | 44.79 | #31 (GPT-5.5) |
| 7 | Gemini 3.8 Flash | 44.76 | #21 |
| 8 | GPT-5.4 (xHigh) | 44.29 | #79 (GPT-5.4) |
| 9 | GLM-5.2 | 42.38 | #73 |
| 10 | GPT-5.3 Codex (xHigh) | 42.38 | #69 (GPT-5.3 Codex) |
| 11 | Claude Opus 4.6 | 35.58 | #63 |
| 12 | Gemini 3.1 Pro (Preview) | 33.81 | #58 |
| 13 | Claude Sonnet 4.6 | 32.21 | #90 |
| 14 | GLM-5 | 24.24 | #141 |
| 15 | Kimi K2.5 | 20.95 | #144 |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=seal-swe-atlas-refactoring · How It Works · Data refreshed daily, snapshot 2026-10-20.