RealBench (Incremental) - Level 1: leaderboard

RealBench (2026) asks LLMs to build a whole Python repository from natural-language requirements and UML system designs, here with incremental generation (file by file until every class exists) on the level-1 repositories (about 260 lines of code). The score is the percentage of the original unit tests that the generated repository passes, higher is better.

Metric: Test pass rate (%), passed tests over all tests of each task, averaged over the level-1 tasks (14 repositories of about 260 lines), with incremental generation (file by file until every class exists); 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: Not forecast. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 235B A22B31.95#633
2Qwen 2.5 Coder 7B Instruct28.71#3976
3Claude Sonnet 4 (20250514)26.8#410
4GPT-4o (2024-05-13)20.45#740
5DeepSeek V317.18#736
6Gemini 2.5 Flash5.64#466

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=realbench-incremental-level-1 · How It Works · Data refreshed daily, snapshot 2026-11-22.