Benchmark / research / guide

Parsimony

Show HN: Parsimony – an open-source benchmark for how much code AI models add

Search and research Source: Hacker NewsScore 3.24 · #27 on the hot list77 points· First listed Oct 15, 2026

Visit the Parsimony website / project page →HN discussion

What is Parsimony

Parsimony is an open-source benchmark that measures how much additional code AI coding models introduce when completing tasks, evaluating model conciseness.

Key features

Measures the amount of additional code introduced by AI coding models when completing tasks
Uses syntactic elements (such as statements, calls, names and literals) as encoding units for counting
Evaluated on public patches from DeepSWE, SWE-bench Verified and SWE-bench Live
Net encoding unit calculation: deleted code is deducted
Supports user submission of public patches for new models

Who it is for

Suitable for AI coding model developers, researchers and developers concerned with long-term code maintainability.

How to get started

Visit https://parsimony.lasu.dev/ to view benchmark results
Go to https://github.com/La5u/Parsimony for the open-source code and detailed methodology
Submit public patches for new models to participate in the benchmark

Compiled from public descriptions on the tool's official website / README / launch page. Pricing and user numbers have not been verified; please check the official website. Some sections were drafted with AI assistance. Written in Chinese and English; other languages are machine-translated.