They publish a benchmark for anyone interested: https://marisa-trie.readthedocs.io/en/latest/benchmarks.html

Summary for storing a list of 3M Russian words:

- ~60x less memory

- ~5-10x slower compared to hashmap