Pac-Bench
Open SourceBenchmark evaluating how well AI models one-shot a Pac-Man game
🐳 Self-Hostable🔓 No Sign-up Required⚡ Traction Signal: 71/100
★17 Stars / Upvotes
💡 What Problem Does Pac-Bench Solve?
Pac-Bench is an interactive evaluation framework designed to test the code generation and spatial reasoning capabilities of LLMs by challenging them to one-shot build a fully working Pac-Man game. It provides developers and researchers with a standardized benchmark to measure frontier model performance on complete game development tasks.
CategoryAI & Machine Learning
Commercial AltStandalone Utility
Self-HostableYes (Docker/Local)
Sign-up RequiredNo (Instant Access)
⚡ Key Capabilities & Architecture
#1One-Shot Game Generation
Challenges AI models to write a complete, playable Pac-Man game in a single prompt.
#2Model Comparison
Compares performance across various frontier LLMs to gauge coding and spatial reasoning.
#3Interactive Playground
Allows users to view and test the resulting generated games directly in the browser.
🎯 Who is this for?
AI researchers, prompt engineers, and developers interested in evaluating LLM code generation capabilities.
Top Alternatives in AI & Machine Learning
Compare other freshly discovered tools and open-source alternatives.
