Why does a long chat cost more per turn over time?

Each API call resends the entire conversation so far as input. By turn ten you are paying to reprocess turns one through nine again, so the input cost of each new turn grows roughly linearly with conversation length.

Is the system prompt billed every turn?

Yes. The system prompt is part of the input on every request, so a long system prompt multiplies across every turn. The tracker adds it to each turn's input count.

How are tokens estimated?

Token counts use a character and word heuristic of roughly four characters per token for English, plus a small per-message overhead for role and structure tokens. It is an estimate, not the exact tokenizer.

How can I keep a long conversation cheap?

Summarize or truncate old turns, use prompt caching for the stable prefix, or move static context into a system prompt that the provider can cache. The tracker shows exactly where the cost is accumulating.

Is anything uploaded?

No. The whole simulation runs in your browser. Nothing you type is sent anywhere.

Conversation Cost Tracker

Conversation cost tracker

The expensive surprise in chat applications is not any single message — it is the way an LLM conversation resends its entire history on every turn. Early turns are cheap; by the twentieth turn each new reply re-bills all nineteen prior turns as input. This tracker simulates that growth so you can see the running total climb and decide when to summarize or truncate.

How it works

A chat API is stateless: to keep context, your app resends every prior message on each request. So turn N’s input cost equals the system prompt plus all turns 1..N-1 plus the new user turn, while the output cost covers only the new assistant reply. The tracker counts tokens for each turn (with a small per-message overhead for role tokens) and accumulates the input and output costs at your chosen model’s rates, updating the meter after every turn you add.

Tips and notes

The system prompt is paid on every turn — keep it tight, and prefer provider-side prompt caching for any long, stable prefix.
Once a conversation gets long, summarizing the oldest turns into a short recap can cut input cost dramatically without losing the thread.
Output tokens are billed only once each (the new reply), but input tokens compound — that asymmetry is why long chats drift toward being input-dominated.