AI models benchmarked with Pac Man

PacBench, a benchmark hosted on GitHub by Jon Clegg, tasks AI models with creating a Pac-Man game in a single HTML page and records each entry's platform, model, generation time, token usage, and cost. Unlike the pelican benchmark on Hugging Face, PacBench's outputs are playable, as reported by Boing Boing on September 30, 2026.

PacBench https://jonclegg.github.io/pacman-bakeoff/ is an unusual benchmark for AI models: "Create a Pac-Man game in a single html page." Each entry shows the platform, the model, how long it took, how many tokens it used, and the cost. And unlike the pelican https://huggingface.co/spaces/victor/pelican-benchmark , we can play the results. — Read the rest https://boingboing.net/2026/09/30/ai-models-benchmarked-with-pac-man.html The post AI models benchmarked with Pac Man https://boingboing.net/2026/09/30/ai-models-benchmarked-with-pac-man.html appeared first on Boing Boing https://boingboing.net .