Using Qwen2.5-0.5B-Instruct to generate news drafts. This model is optimized for CPU efficiency and instruction following.