Glossary
/
Performance

Token Efficiency

How few tokens an AI uses to finish a task, which drives its cost.

Definition

Token efficiency is how few tokens (the units of text an AI model reads and writes) an agent needs to finish a task. Fewer tokens means lower cost per conversation. A knowledge graph helps by sending the model only the facts it needs, not whole documents, cutting token use by up to about 60%.

Key points

<ul><li>Example: instead of sending the model a full catalogue PDF, the agent sends only the price, stock and pack size for one SKU.</li><li>Small savings add up when an agent handles thousands of conversations.</li><li>Sending less to the model can also reduce latency. Measure it alongside accuracy, because a cheap wrong answer isn't a saving.</li></ul>