Token efficiency: Cut prompts to 250 tokens
Short prompts of 250 tokens keep models in peak form while longer inputs cause measurable degradation in output quality and speed.
Marcus Chen
13 min read