Token counter for Claude, GPT and Gemini
Paste anything and watch it counted against every major model as you type. Counts for GPT-4o, GPT-4.1, and o3 are exact in your browser. For Claude and Gemini, one click asks Anthropic's and Google's own tokenizers, so the number you see is the number the API bills.
0 words · 0 characters
Context windows compared
A context window is how much a model can read at once. The bars fill as you type above, so you can see exactly how much room your text takes in each model.
| Model | Your text | Context window | Fit | Max output | Input $/M tokens |
|---|---|---|---|---|---|
| Claude Opus 4.8 Anthropic in NodePad no long-context surcharge | — | 1M ~750K words | 128K | $5 | |
| Claude Sonnet 5 Anthropic in NodePad | — | 1M ~750K words | 128K | $3 | |
| Claude Haiku 4.5 Anthropic | — | 200K ~150K words | 64K | $1 | |
| GPT-5.5 OpenAI in NodePad very long prompts bill at a higher tier | — | 1.1M ~788K words | 128K | $5 | |
| GPT-4.1 OpenAI | — | 1.0M ~786K words | 33K | $2 | |
| GPT-4o OpenAI | — | 128K ~96K words | 16K | $2.5 | |
| o3 OpenAI | — | 200K ~150K words | 100K | $2 | |
| Gemini 3.1 Pro Google in NodePad $2 up to 200K-token prompts, $4 beyond | — | 1M ~750K words | 64K | $2 | |
| Gemini 2.5 Pro Google $1.25 up to 200K-token prompts, $2.50 beyond | — | 1.0M ~786K words | 66K | $1.25 | |
| Gemini 2.5 Flash Google | — | 1.0M ~786K words | 66K | $0.3 | |
| Grok 4.3 xAI max output not published by xAI | — | 1M ~750K words | — | $1.25 | |
| Llama 4 Scout Meta 10M advertised by Meta; most hosts serve 128K to 1M | — | 10M ~7.5M words | — | — | |
| DeepSeek V4 Flash DeepSeek | — | 1M ~750K words | 384K | $0.14 |
Windows and prices verified against provider documentation on 2026-07-07. The price column is each model's API input rate; the line under your live count is what sending your text once would cost at that rate.
You just managed context by hand. NodePad is a canvas for doing it on purpose: fork side questions, keep every path, and choose exactly what each model sees.
Try NodePad freeToken questions, answered
What are tokens in AI?
Tokens are the pieces a model actually reads: chunks of characters, roughly four characters or three quarters of an English word each. Context windows are measured in tokens and API pricing is billed per token, which is why counting them matters.
How many tokens is a word?
English prose averages about 1.3 tokens per word, so 100 tokens is roughly 75 words. Code, other languages, and unusual formatting tokenize less predictably, which is why a real count beats the rule of thumb.
Why are some counts estimates?
Anthropic and Google do not publish their tokenizers, so no browser-side tool can count them exactly; the exact-count button asks their own counting APIs instead, which is the only way to get the true number. OpenAI publishes encodings for GPT-4o, GPT-4.1, and o3, so those are exact locally. GPT-5.5's encoding is not published yet, so it is estimated like the rest.
How many tokens does Claude Pro include?
Anthropic does not publish a fixed token allowance for Claude Pro. Limits are usage-based: they scale with demand, reset every few hours, and long conversations spend the allowance faster because the whole conversation is resent with every message. Counting your context is the best way to stretch it.
How much do tokens cost?
In the chat apps (ChatGPT Plus, Claude Pro, NodePad) you pay a subscription, not per token. On the APIs, every model has a per-million-token price. The table above shows each model's input price next to your live count.
What is ChatGPT's token limit?
It depends on the model behind it: GPT-4o reads 128,000 tokens at once, while GPT-5.5 and GPT-4.1 take roughly a million. The ChatGPT app manages this silently, so long conversations quietly fall out of the window. Paste your text above to see how much of each window it uses, or see every model's limit in the context-window comparison.
Do long conversations use more tokens?
Yes. Chat apps resend the whole visible conversation with every message, so message thirty carries the previous twenty-nine with it. That is why limits feel tighter late in a long thread, and why tools that let you control the context, rather than resend all of it, stretch the same allowance further.