You can use spacy for english tokenization or you can use their neural model. The neural model will generally do better, especially sentence segmentation, but will be slower.