An estimate of context use. Not a tokenizer.
0 tokens · 0% of 128,000 · Fits An estimate from 0 characters, not a tokenizer count.
This text may exceed the selected context window.
▸ More about
Uses of context window
- Checking a long draft before pasting it into a model with a published window
- Seeing that a character count and a token estimate are different numbers
- Reading which vendor page a window was copied from
context window pitfalls
- Treating the estimate as the model's own tokenizer
- Ignoring the date on the source note after a model update
- Expecting a price
context window questions (4)
Is the token count exact?
No. Estimated tokens are the number of characters divided by 4, rounded up. A model tokenizer can split the same text differently. The page says estimate on purpose.
Where do the context windows come from?
Each model row names the vendor page it was copied from, and the date that page was read. ToyTools does not guess a window that is not on that list. Windows change, so the date matters.
Does this pick a model for me?
No. You choose a model. The page only compares the estimate with that model's published window. It does not rank models or call one.
Is my text uploaded?
No. The estimate runs in your browser. Runs entirely on your device. Nothing is uploaded. Reloading the page clears the text.