Smaller Language Models for Enterprise

Preference for models between 10-70 billion parameters over larger models due to inference cost efficiency