The Pac-Man Benchmark: one prompt, four models, wildly different HTML results
neverbyte · reddit · 2026-11-11
A Reddit user tested multiple models with the same prompt — "create Pac-Man in HTML in a pixel art style". Per the comparison image, Qwen 3.8 Next Flash, DeepSeek 1017, and Opus 5.5 all produced mazes close to the original layout, while Fable 5.2 output a brick-walled, structurally off version — and the author says that's all Fable would give even after extra prodding. These same-prompt showdowns have become a popular quick way to compare front-end generation quality.
More from Fun
- "I Feared AI Would Kill Human Creativity — Then I Saw TikTok Grocery Haul Videos" — signulll · 2026-11-11
- AI drama: cheating scandal followed by a studded leather jacket appearance — "you come at the king you best not miss" — inductionheads · 2026-11-11
- Programmer-as-archaeologist meme: digging through code history to find Carl's ruins and evil Paul's damage — zetalyrae · 2026-11-11
- Tesla FSD V14.3.13 spares a bunny's life, owner shares the save — justin_hart · 2026-11-11
- Microsoft finally adds 'Don't ask me again' to Sticky Keys dialogs after three decades — sanjaykalra · 2026-11-11
- $200 Claude Ultra plan drained in one night: usage screenshot shows weekly limits maxed — ezshine · 2026-11-11