Chat with a PDF, Word, Excel or PowerPoint file

What you can attach

Quad reads text from the files most people actually have. In the chat composer you can attach: PDFs, including long papers and reports. Word documents (.docx and .doc) and rich text files. Excel spreadsheets and CSV files. PowerPoint presentations. Plain text, Markdown, JSON and many code files. Images, such as a photographed page or a screenshot of a chart, on models that accept image input. You can attach up to eight files to one message. Documents can be up to 10 MB each and images up to 8 MB. For a book-length PDF, upload it by chapter rather than as one file.

How to get a reliable answer

Ask for structure first: a summary by section, key terms, dates and numbers, or a table of obligations. Ask the model to separate what the document states from what it infers, and to quote the passage and page it relied on. Request specific extractions rather than vague summaries, for example every deadline, price and renewal term in a contract. Send the same question to a second model. Where the answers differ, open the page and read it yourself. Check every quotation and number against the original. Models can misread a table or attribute a statement to the wrong party. Switching models inside the same conversation is built for exactly this: summarize with one model, challenge that summary with another, and keep both answers beside the source.

Saving files and keeping them organized

A temporary attachment is used for that reply only, and you reattach it after reloading the page. To reuse documents, save them to your Library or keep them in a project, which groups conversations and source files for one body of work so new chats can start with the right material. Free (File storage: 250 MB; Objects: 250). Go (File storage: 2 GB; Objects: 2,000). Pro (File storage: 10 GB; Objects: 10,000). Ultra (File storage: 50 GB; Objects: 50,000). Max (File storage: 250 GB; Objects: 250,000). The Free plan needs no credit card and includes 100 credits a month. Longer documents use more credits because they put more text in the request, and models with larger context windows can take more of a document at once.

Limits to know before you rely on it

Scanned PDFs: Quad extracts selectable text. A PDF that is only page images may return little or nothing, so run OCR first or attach the pages as images to a model that accepts them. Very long documents are truncated when their text is extracted, and the model is told that the text was cut. Split long files and ask about one part at a time. Tables and layout can lose structure in extraction. For a critical table, ask the model to restate it and compare it with the page. Files attached temporarily to a message are sent to one model at a time, so side-by-side comparison of an attached file is not supported. Ask one model, then switch the conversation to another. Incognito chats turn off new uploads. Use a normal chat when you need to attach files. A document is not legal, medical or financial advice. Use the output to prepare your questions for a qualified professional.

What happens to your files

Quad does not use your conversations or files to train its own models, on any plan. Files and chats are encrypted in transit and at rest, and you can delete them yourself. The text of a file goes to the model you chose in order to answer, so the same rules apply as for any prompt: most models run on zero-data-retention provider routes, and models that do not ask you to confirm first. See the private AI chat page for the full picture, and keep confidential documents out of any tool your contract does not allow. Dedicated notebook tools such as Google's NotebookLM are built around source-grounded notes and are worth trying if that is all you need. Quad is the better fit when the document is one input among web research, drafts, images and projects.

Frequently asked questions

Can I chat with a PDF for free?

Yes. The Free plan needs no credit card and includes 100 credits a month and 250 MB of file storage. Longer documents use more credits.

What file types can I attach?

PDF, Word (.docx and .doc), RTF, Excel and CSV, PowerPoint, plain text, Markdown, JSON, many code files and images. You can attach up to eight files per message, documents up to 10 MB and images up to 8 MB.

Does Quad read scanned PDFs?

Quad extracts selectable text, so a scanned PDF that is only images may return little or no text. Run OCR on it first, or attach the pages as images to a model that accepts image input.

How long can a document be?

Very long documents are truncated when text is extracted, and Quad marks the cut. For book-length material, split it into chapters and ask about one at a time.

Are my uploaded files used to train AI?

No. Quad does not use your conversations or files to train its own models, on any plan. Files are encrypted in transit and at rest, and you can delete them yourself.

Can I compare how two models read the same PDF?

Not at the same time with a temporarily attached file, which goes to one model at a time. Ask one model, switch the conversation to another and ask again, then compare the answers.