About ASCII Bench

¯\_(ツ)_/¯

What is ASCII Bench?

ASCII Bench is an A/B voting platform designed to evaluate and compare the quality of ASCII art generated by different AI models. Users are presented with pairs of ASCII art outputs created from the same prompt and vote for the one they prefer.

Through community voting, we build a leaderboard that ranks which AI models produce the best ASCII art. This helps researchers and developers understand model capabilities in creative text-based visual generation.

How do rankings work?

Each AI model has a single ELO rating that represents its overall performance across votes. When you pick a winner in a head‑to‑head matchup, the winner gains points and the loser loses points based on how expected the outcome was.

We use a dynamic K‑factor: new models start with larger rating changes so they find their true level quickly, and as a model accumulates votes the changes shrink to keep the leaderboard stable.

Creator

( ͡° ͜ʖ ͡°)
Jason Botterill

Jason Botterill

Owns 10%

( ͡ಠ ͜ʖ ͡ಠ)
Chris

Chris

Contributor

(づ｡◕‿‿◕｡)づ
Angel

Angel