NITR (API): leaderboard
Metric: API-mode pass rate (%) over NITR's 21 C++ repository probes in nine maintainability dimensions: an edit passes only if the hidden functional tests and the structural maintainability oracle both pass; one run per probe; direct inference without scaffolding, each step answered with the relevant prior context; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 38.1 | #54 |
| 2 | GPT-5.4 | 33.33 | #76 |
| 3 | Gemini 2.5 Pro | 28.57 | #145 |
| 4 | Claude Sonnet 4.5 | 28.57 | #138 |
| 5 | Claude Opus 4.6 | 28.57 | #60 |
| 6 | GPT-5 Mini | 28.57 | #176 |
| 7 | Claude Opus 4.5 | 28.57 | #79 |
| 8 | GPT-5.3 Codex | 28.57 | #68 |
| 9 | Claude Sonnet 4.6 | 23.81 | #85 |
| 10 | Qwen3 Coder | 23.81 | #285 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=nitr-api · How It Works · Data refreshed daily, snapshot 2026-10-11.