RepoMod-Bench: leaderboard
Metric: Pass rate (%): share of the hidden tests passed by the translated repository (0 when it does not build), averaged over the 21 RepoMod-Bench source-to-target translation tasks (11,616 tests; prompts forbid subprocess wrappers, FFI calls to the original binary and vendoring); one autonomous agent run per task with a 4-hour limit; higher is better. Source: arxiv.org. Saturation forecast: Around July 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 43 | #105 |
| 2 | Claude Opus 4.5 | 42 | #79 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=repomod-bench · How It Works · Data refreshed daily, snapshot 2026-10-11.