The models can as well be found in the SF download area of the OpenNLP project.
It is not recommended to use those models, they are suitable to get an impression of what OpenNLP can do. To get good results it is necessary to train models on your own data, or mix our data with existing training data. OpenNLP has built-in support to be trained on a number of corpora, if you would llke to process English news content you should consider training on the OntoNotes corpus. There is a section in the manual listing all the supported corpora and how to train on them. Jörn
