In this space you can compare some of the features of code generation models:
* Pretraining datasets
* Architecture
* Model evaluation
You can also test their code generation capacities through prompting.