Models, Tokens, and Cost

Not all language models behave the same way. Some models are: - smaller - faster - cheaper Others are: - larger - more capable - better at reasoning - more expensive Choosing the right model depends on the task. Simple classification tasks may work well on smaller models. Complex reasoning tasks may require larger models.

**Tokens** Models process text as tokens. Tokens are small chunks of text: words, parts of words, and punctuation. Both input tokens and output tokens contribute to cost. Long prompts and long responses increase usage and latency.

**Context Windows** Models also have context limits. If a prompt becomes too large, the model may: - truncate information - lose important context - fail to follow instructions reliably Efficient prompt design matters because real systems operate under cost limits, latency limits, and context limits.