AI model leaderboard
From writing code to modifying repositories, testing whether a model can actually build software.
3 evaluations · 3 organisations · updated 11/13 14:05
Codingtop 30
Ranked on the evidence several public evaluations share
Each board shows at most 30 models
The consensus index is not an accuracy rate; tied scores keep the order the shared evidence decides.
How to read this board
Aggregates several public evaluations, and different models have different coverage. Prices play no part in the ranking, a missing evaluation does not count as zero, and the index is not an accuracy rate.
How it is calculatedAbout prices
API prices come from the providers' own sites, per million tokens. Dollar prices are converted to CNY at the rate of 2026-11-12. The cached price is the input price on a cache hit; cache writes, storage and subscriptions are extra.