What's in the other answer is correct (and I can't comment due to reputation).

There is an important caveat though:

If you write to a graphmart, which is the Anzo abstraction for a graph, the write goes to a linked data set called journal. The triples you write will not be available in the graph automatically, unless you trigger a graphmart refresh, which essentially means running the whole ingest pipeline again. This adds at least a few seconds of delay, depending on the amount of data, where the whole graph is unavailable.

If your usecase relies on OLTP with a sufficiently high number of writes, and n seconds of lag are a no go, you should consider a different platform.

For persisting data in an OLTP way, Anzo uses those journals, which actually are an instance of Blazegraph running.

(This is the behavior on the Anzo 5.3 platform, it might have improved since.)