checkpoint
A saved, trained version of a model, ready to run without retraining.
Training a model produces a checkpoint. It is a snapshot of everything the model has learned so far, saved as a file. The file can be loaded and used straight away. It can also be trained further later.
Think of it as
Like a saved game file that lets you pick up exactly where you left off, rather than starting over.
Example
A project publishes three checkpoints: one for English, one for many languages, and one tuned for business decisions. You download the one that fits your work and start using it the same day, without training anything yourself.