CodeScope - Code Understanding: leaderboard
The Code Understanding subtask of CodeScope (arXiv) — reasoning about program behavior and semantics; score.
Metric: Score. Source: haitianliu22.github.io. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-3.5 | 49.2 | #818 |
| 2 | GPT-4 | 46.93 | #438 |
| 3 | starcoder | 44.04 | #1577 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=codescope-code-understanding · How It Works · Data refreshed daily, snapshot 2026-10-20.