It's slow because you are reading into an in-memory store (RDF.Model() default) which has no indexing. So it gets slower and slower. The parsing of N-Triples does stream from the file, it never sucks it all into memory.
See the Redland storage modules documentation for an overview of the storage models. Here you probably want storage type
'hashes' and hash-type
memory.
s = RDF.HashStorage("abc", options="hash-type='memory'")
model = RDF.Model(s)
(not tested)