SEAL - SWE Atlas - Codebase QnA: leaderboard
Scale SEAL evaluation of LLM ability to answer questions about large codebases, testing code comprehension and navigation.
Metric: Score. Source: scale.com. Status: years away from saturation. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GLM-5.2 | 48.12 |
| 2 | GPT-5.6 Sol (xHigh) | 46 |
| 3 | GPT-5.5 (xHigh) | 45.43 |
| 4 | GPT-5.4 (xHigh) | 36.3 |
| 5 | DeepSeek V4 Pro | 27.15 |
| 6 | Muse Spark | 24.2 |
| 7 | GLM-5 | 20.5 |
| 8 | Gemini 3.1 Pro (Preview) | 13.5 |
| 9 | Kimi K2.5 | 13.1 |
| 10 | MiniMax-M2.5 | 10.3 |
| 11 | Gemini 3 Flash | 8.2 |
Interactive version: theaggregate.ai/benchmark?slug=seal-swe-atlas-codebase-qna · How It Works · Data refreshed daily, snapshot 2026-09-05.