Assessing The Potential Of Mid-Sized Language Models For Clinical QA