---
title: What one message costs — AI hygiene, lesson 1
description: Every message sends the whole window to the model again, so one message in a 100,000-token window reads $0.40 of tokens on Opus 5.5, and in 6,000 tokens $0.024.
url: https://instibrain.ai/guides/ai-hygiene/what-one-message-costs
updated: 2026-11-08
site: https://instibrain.ai/llms.txt
---

lesson 1 of 8 · ai hygiene

# What one message costs.

Every message sends the whole window to the model again. The fuller the window, the more each message costs.

## new words

**Window**
Everything the model reads for one message: its instructions, the files it read, and the whole conversation so far. Also called the context window.

**Input tokens**
The tokens the model reads. On Claude Opus 5.5 they cost $4 per million.

**Output tokens**
The tokens the model writes. On Opus 5.5 they cost $20 per million.

**Ctx**
On a computer linked to instibrain, the share of the window in use, shown in the status line under Claude Code's input. Without instibrain, `/context` shows it.

## the lesson

A model keeps nothing between messages. Each time you press Enter, your AI sends the whole window again, and the model reads it from the first token: How a model works, lesson 1. The new message is the small part at the end.

So the price of a message is mostly the window, not your question. At Opus 5.5's list price of $4 per million input tokens, before the cache of lesson 3:

| Window | Read for one message |
| 6,000 tokens | $0.024 |
| 60,000 tokens, Ctx 6% of 1,000,000 | $0.24 |
| 100,000 tokens | $0.40 |
| 500,000 tokens | $2.00 |

The answer is cheap beside it: 500 tokens written at $20 per million is $0.01. In a 100,000-token window, the reading is 40 times the writing.

On a Claude plan you see no bill. Anthropic names the conversation's length among what uses up a plan's limits, so the same window spends your limits instead. instibrain's status line shows those limits too, as 5h and 7d.

The Ctx number is how full the window is, and so how much each message now costs. Read it before you type.

The price of reading the window once, by its size, drawn to scale. Your question is a few dozen tokens of it.

## try it

In Claude Code, type this and press Enter:

```
/context
```

## you should see

- A table of what fills the window: the system prompt, tools, memory files such as CLAUDE.md, messages, and the free space.
- A total close to the Ctx number in the status line, if your computer is linked to instibrain.

---

instibrain is made by WebWork AI, Inc. Everyone uses the same models · the difference is what yours knows.
