Parsimony – an open-source benchmark for how much code AI models add
Summary
Parsimony introduces an open-source benchmark that measures how much code AI models add by counting coding units (statements, calls, comparisons, names, literals) and computing net additions. Models are ranked by mean net units added on solved tasks, with tests sourced from SWE-bench; the leaderboard, data sources, and methodology are documented. The project invites community participation and provides a browser-based benchmark workflow.