AIBugBench

Deterministic, local AI code benchmarking — sandboxed and reproducible.

Compare models on the same four tasks. Score across seven dimensions. No network. No vibes. Just receipts.

Get started How scoring works

Pick your path

• New users
  Install, run your first benchmark, read the scorecard.
  Getting Started

• Model authors
  Add a model folder and wire outputs.
  Developer Guide

• Power users
  CLI flags, artifacts, diffs, comparisons.
  User Guide

• Security first
  Sandbox, audit, and guardrails at a glance.
  Security

Documentation

Core Guides

• Getting Started - Setup and first benchmark run
• User Guide - Running benchmarks and interpreting results

Understanding the System

• Scoring Methodology - How the 7-category scoring works
• Troubleshooting - Common issues and solutions
• Sabotage Notes - Debug patterns and test hazards

Project Information

• Contributing - Development workflow and contribution guidelines
• Code of Conduct - Community guidelines
• Security - Security policy and reporting
• License - Apache-2.0 license information
• Roadmap - Project roadmap and future plans
• Release Notes - Version history

Developer & Internals

Developer Guide Architecture API Reference Internals Overview

Validation Scripts Submissions Template

Quick start

See Getting Started for installation and your first run.