How does similarity relate to topic number?

[Nicolas Posner]

Gensim beginner here. Is there any reason to expect that similarity between the same documents and queries in different models should vary because of different numbers of topics, or is variance there just because of randomness in the model training itself?