—
per month
—per conversation
—per year
—vs human agents

Cheapest models for this chatbot

Same traffic, every model ranked by monthly cost.

ModelCost / monthPer conversation
⚠️ Estimate using reference prices (June 2026) and list rates. Real bills vary with caching, tools, retrieval, batching, region and tiers. Long chats cost more than people expect because history is resent every turn — that's modelled here.

Why chatbot cost grows faster than message count

The trap with chatbots is resent history. To stay coherent, most bots send the whole conversation so far with every new message. So the 1st reply pays for one message of context, the 10th pays for nine. Total input tokens grow roughly with the square of the conversation length — which is why a few very long chats can dominate your bill. Trimming or summarising history, capping conversation length, and using a smaller model for routine turns are the biggest levers. Try the "Summarised history" option above to see the difference.

Comparing to humans? A short chat on a small model often costs a fraction of a cent — far below a human ticket. But a long chat on a frontier model with full history can cross into dollars and beat a human's cost. The break-even above tells you where you stand. Building something bigger than support? Use the AI app cost estimator or compare models head-to-head on the AI API cost calculator.