RealBench (RAG) - Level 3: leaderboard
Metric: Test pass rate (%), passed tests over all tests of each task, averaged over the level-3 tasks (20 repositories of about 1,340 lines), with retrieval-augmented generation (file by file, each conditioned on the earlier generated files it imports); RealBench repository-level Python code generation from natural-language requirements plus UML package and class diagrams (61 real repositories in four size levels), one greedy solution per task through the official APIs; higher is better. Source: arxiv.org. Saturation forecast: Around 2033. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4 (20250514) | 19.49 |
| 2 | GPT-4o (2024-05-13) | 19.1 |
| 3 | DeepSeek V3 | 17.37 |
| 4 | Qwen 2.5 Coder 7B Instruct | 17.08 |
| 5 | Gemini 2.5 Flash | 14.91 |
| 6 | Qwen 3 235B A22B | 7.16 |
Interactive version: theaggregate.ai/benchmark?slug=realbench-rag-level-3 · How It Works · Data refreshed daily, snapshot 2026-10-07.