Show HN: Better Than 4 But Not 5 – An LLM model ordering challenge

wspittman.github.io

1 point

a year ago

This simple game challenges you drag-and-drop a set of LLM models into the correct order (release date, input cost at launch, LM Arena score) using just their name.

It's no secret that LLM model names are a bit of a mess, but when OpenAI decided to backtrack from 4.5 to 4.1, well, I just couldn't let it go.

This was my first attempt at vibe coding something with VSCode Copilot Agent (w/ Claude 3.7 Sonnet). The agent got me like 95% functionality, but the moment I needed to crack open the code myself I couldn't handle the mess and had to clean it up. I am bad at letting go control enough for vibe coding. If anyone else has had that problem and gotten over it, I would love some helpful tips.

I collected the data primarily by digging through Simon Willison's blog archives and snapshotting LMArena's leaderboard. If you spot any inaccuracies in the model data, I'd appreciate corrections with sources!

Nothing fancy in the tech here, just a vanilla HTML/CSS/JS site hosted on GitHub Pages. GitHub link (MIT License): https://github.com/wspittman/BetterThan4ButNot5