Tokens per Request: Words, Context & Multi-Turn Chats
This lesson is part of the course
In this lesson
You estimate the tokens of one Fernhill Assist support reply from word counts, then see how a six-turn chat multiplies the input tokens because the history is sent again every turn.
Buy the course once for $9.99 to unlock every lesson and lab, the final exam and the certificate – lifetime access, no subscription.
Lessons in this course
- Tokens per Request: Words, Context & Multi-Turn Chats Locked
- Cost per Request & per 1,000 Requests with a Price Table Locked
- Monthly Cost from Usage Tiers: Light, Regular & Power Users Locked
- Prompt Caching & Batch Discounts: When Do They Pay? Locked
- Model Routing: Blended Cost vs Quality Locked
- RAG Cost: Embeddings, Retrieved Context & the Monthly Bill Locked
- Cost per Resolved Ticket vs Human Cost: the Break-Even Locked
- Pricing an AI Feature: Gross Margin vs Usage Locked