Best LLM for Coding: Ranked by People Who Judge the Code

It uses the coding category of Arena's blind votes. Voters see two anonymous answers to a programming question and choose the better one. The rating comes from how often each model wins.

No. People judge the answer they read. Benchmarks such as SWE-bench Verified run the code against real project tests, which is a different and stricter kind of check. Looking at both gives a fuller picture.

Not always. A leaderboard compares models on one kind of test. Price, speed, privacy rules, the apps you already use and the type of work you do all matter. Use the rank as a starting point, then try two or three models on your own tasks.

A model appears only when an open source has published results for it. New models can take days or weeks to be measured, and models with no published score are left out rather than guessed.

Ratings come from the open leaderboard dataset published by Arena (arena.ai), licensed CC BY 4.0. We keep the latest published ratings, match models to our other sources and rank them. Arena does not endorse this site. Prices and context windows come from the open-source LiteLLM price list (MIT licence), which records each model maker's own price page. Prices change often, so confirm on the maker's site before you buy.

We check our open data sources every day and show the date of the data at the top of the page. A model only moves when its source publishes new results, so a number can stay the same for weeks.