Parsimony
Parsimony is a benchmark that ranks AI coding models by how much extra code they add to fix bugs, for developers comparing models.
What it is good for
- Comparing how much code different AI coding models add
- Judging which AI coding agent makes smaller changes
- Checking whether a coding model's fixes bloat a codebase
At a glance
- Who it is for
- Setup
- Not stated
- Cost
- Free and open source
- Type
- Other
- Tasks
- Works with
- SWE-bench Verified, mini-SWE-agent, Claude Code, Codex
Momentum
- Rank this week
- Not ranked yet
- Best rank
- None yet
On Hacker News
- Show HN: Parsimony, an open-source benchmark for how much code AI models add 77 points, 19 comments, Oct 13, 2026
Official links
Setup steps are on the official pages. We do not repeat them here, since they change.
Questions
- Is Parsimony free?
- Yes. It is free and open source.
- What does Parsimony work with?
- According to its own documentation, it works with SWE-bench Verified, mini-SWE-agent, Claude Code, and Codex.
- What is Parsimony good for?
- Comparing how much code different AI coding models add, judging which AI coding agent makes smaller changes, and checking whether a coding model's fixes bloat a codebase.
Similar agents to Parsimony
- Understand-Anything
Understand-Anything is a plugin that turns a codebase, knowledge base, or docs into an interactive graph you can explore and ask questions about, for people new to a project.
- OmO
OmO is a coding agent app that turns a single prompt into finished software work for developers and builders who want agents to handle multi-step jobs.
- AgenticSeek
AgenticSeek is a self-hosted agent app that browses the web, writes code, and plans tasks for people who want to keep their data on their own machine.
- GitDiagram
GitDiagram is an MCP server and web tool that turns a GitHub repository into an architecture diagram or narrated video for people who need to understand a codebase.
- MiMoCode
MiMoCode is a terminal-native AI coding assistant that reads, writes, and runs code for developers working in a command line.
- LoopX
LoopX is a local-first control plane that keeps long-running agent work on goals, decisions and evidence across sessions and tools for engineers and researchers.