PasteBudget
Private in-browser context calculator

Your context window is not your paste budget.

See what actually fits after prior messages and answer space. If it’s too large, turn it into safe, paste-ready parts in one click.

Real API limitsLocal file readingNo prompt upload
Check my real budgetUse the demo below

The 200K context trap

Claude Haiku 4.5

Won’t fit
Published context window200,000
Already used by the chat−191,850
Protected for the answer−8,000

Actual room for new material

150

Example needs

≈161

PasteBudget catches the 11-token overflow.

Then it makes 2 safe parts without uploading the text.

Check the whole request

Find out if it fits. Fix it if it doesn’t.

Paste the real prompt, account for the conversation already used, and protect enough room for a useful answer.

Your text never leaves this tab
1
Add what you’re about to send
Prompt, instructions, document, transcript, or source code.
0 characters0 lines0 words0 min read
2
Set the real budget
Model window minus answer space and prior conversation.

OpenAI API · limits checked September 2, 2026 · official source

16K

PasteBudget subtracts this from the window instead of pretending the entire context is available for input.

Optional

3 · Paste decision

Waiting for your prompt

Add your material to see the budget the model can actually use.

Tokens0/ 1,034,000
0% used1.03M available

Reference count

0

Answer reserve

16K

Existing context

0

Model window

1.05M

Make paste-ready parts
When material is too large, split it locally at readable boundaries.

Chunking happens only in this tab. Labels get 24 tokens of reserved room.

The method

One useful verdict, with the caveats visible.

A raw context-window number is not your real input budget. PasteBudget subtracts the space you need for an answer and anything already in the conversation before deciding whether new material fits.

01
Measure

Count the pasted text and local files together, off the main browser thread.

02
Budget

Reserve response space and account for earlier messages before the verdict.

03
Split

Turn long material into labeled parts at paragraph or sentence boundaries.

Current references

Model limits with receipts.

Presets are sourced from provider documentation. Limits can change, so every row links back to the current source.

GPT-5.6OpenAI API1.05M128KOpen
Claude Fable 5.1Claude API1M128KOpen
Claude Opus 5Claude API1M128KOpen
Claude Sonnet 5Claude API1M128KOpen
Claude Haiku 4.5Claude API200K64KOpen
Gemini 3.8 FlashGemini API1.05M65.5KOpen
Gemini 3.7 FlashGemini API1.05M65.5KOpen
Grok 4.20xAI API1MNot statedOpen

Practical guides

Use the limit well, not just fully.

Token basics

How many tokens is 1,000 words?
A practical range, why language matters, and when to count instead of estimate.
Read guide

Long documents

How to split long text for an AI chat
Choose a safe chunk size, preserve structure, and carry context between parts.
Read guide

Model limits

Context window vs. output limit
The two numbers are related, but they are not interchangeable.
Read guide

Good to know

Plain answers.

PasteBudget is designed to be conservative about both context and privacy.

PasteBudget

A small, private utility for making better use of finite context. No account, no backend, no prompt collection.

PrivacyToolFiles stay local