Understanding tokens

What tokens are, the difference between per-conversation and monthly limits, and how usage is tracked.

A token is a small chunk of text, very roughly a few characters. Each message you send and each answer you get back uses some tokens. ElegantInsights gives every conversation a token budget so a single chat can't run away, and your plan includes a monthly allowance shared across all your conversations.

What counts

You're charged in full for the new content each turn: your message and the assistant's reply. Everything else is cached: the conversation so far, and the large shared instructions that power every chat, are both re-read each turn at only a small fraction of full price, never re-sent in full. The shared instructions are also shared across all users, so you're never charged to set that cache up.

How a conversation grows

Drag to see roughly how many tokens a conversation uses as it goes on. You pay full price for each new message; re-reading the conversation so far is cached, so only a small fraction counts.

1 turn100 turns

Message turns

Est. tokens used

50,343

input + output

Of the 120k default

42%

~23 turns fit

Estimates for a typical text conversation. Large file attachments use a lot more per message, so they reach the limit sooner.

A typical text conversation uses a few thousand tokens per turn, so the 120k default comfortably covers a normal back-and-forth. Large file attachments use far more per message.

Two kinds of limit

Per-conversation limit

Caps how many tokens a single chat can use. Default 120,000; you can set 20,000 to 1,000,000.

Monthly allowance

Your plan's total tokens for the billing period (3 million on the standard plan), shared across every conversation.

The per-conversation limit keeps a single chat focused; the monthly allowance is your overall budget. Raising a chat's limit lets that chat run longer; it still draws from your monthly allowance.

How a chat's limit is decided

1st

Per-chat override

A limit set for one specific conversation. Wins if present.

2nd

Your global preference

Your default for all chats, from AI settings or the in-chat control.

3rd

Default, 120,000 tokens

Used when neither of the above is set.

The global preference is set in AI settings or via the in-chat control. Try different limits and see the turns each allows:

Per-conversation token limit

20,0001,000,000

Chosen limit

120,000

Est. message turns

~23

typical text chat

75% warning

90,000

first nudge

90% warning

108,000

near limit

Turn estimates are a guide for a typical text conversation; large attachments use more per message. Your plan's token allowance is what actually applies.

Caching: why re-reading the conversation costs only a small fraction.
Tokens in chat: the in-chat budget bar and raising a limit.
Use fewer tokens
Increase your tokens