Comparing Llama3 and DeepSeekR1 on Biomedical Text Classification Tasks