A token is a unit chunk of text. Perplexity AI reads and writes text one token at a time. Perplexity remembers all previous questions up to a context window, which is measured in tokens.

By default Perplexity reads at least 4000 tokens per question but it can read many more with file upload. Longer pasted text is converted to a file automatically.

To unlock longer context windows, please subscribe to Perplexity Pro. With a Pro subscription, file uploads can be read with a context window of at least 32000 tokens by GPT-4 or Claude 3. In addition, GPT-4 and Claude 3 can write up to 4000 tokens at a time.