PLATE DOCArena

Race models on your actual codebase

Benchmarks tell you how a model does on someone else's problems. Arena runs the same prompt through several models against your repository, so you can see which one understands your code.

Same prompt, several models

Pick the contenders, send one prompt, and compare the output side by side. Mix local and cloud models in the same race to find out whether the paid one is actually worth it for your work.

It answers a real question

The useful question is not which model is best in general, it is whether a 7B on your own GPU is good enough for your codebase. Arena is how you find out before paying for anything.

What it does not do

Arena is on a 14-day trial and then part of the Pro plan. Local models remain free with or without it.