Benchmark

NL2Repo-Bench

Natural-language-to-repository generation.

Score history →

Model scores

#ModelScore
1Claude Opus 575.3%
2DeepSeek V4.1-Flash 101768.2%
3DeepSeek V4.1-Flash 091065.4%
4DeepSeek V4-Pro 081361.5%
5GLM 5.358%
6Kimi K358%
7GPT 5.6-Sol56.8%
8DeepSeek V4-Flash 073154.2%
9MiniMax M342.1%