Computational linguistics

A model assigns nearly equal probabilities to many possible next tokens in one context. What does this distribution indicate about predictive entropy for that context?