Upload a PDF to Vibe for Automatic Extraction

Updated November 2, 2026 — Le Chat is now Vibe. As of May 28, 2026, Le Chat has been rebranded as Vibe (still at chat.mistral.ai): a single agent for work and code, carrying over your conversations, settings, and subscriptions. The legacy Le Chat experience remains available in Chat mode; in Work mode, agents have become skills (invoked with @), memories have become the knowledge base, and deep search has become the Deep Research skill. Plans are now called Mistral Free, Pro, Team, and Enterprise. This lesson uses the name Vibe; note that some menus described here may have changed names or locations.

Your First Document in Vibe

Mistral Vibe lets you upload a PDF or document image directly to extract its content. Whether you have an activity report, an invoice, a contract, or a scientific paper, the process is always the same: you upload, you ask your questions, and you get answers. This lesson guides you step by step, from the upload action to the types of questions that get the most out of the document.

The upload procedure is intentionally simple:

1. Click the attachment icon (the paperclip) in Vibe
2. Select your PDF file from your computer
3. Add your instruction or question
4. Send — Vibe processes the document and responds

You can also drag and drop the file directly into the conversation area, which achieves the same result. If your document is an image—a photo of a form, a JPEG scan, or a screenshot of a document—the procedure is identical: Vibe automatically detects it as a document and activates the appropriate OCR processing. Regarding formats, PDF is the most common, including multi-page files; JPEG, PNG, and WebP images work well for scanned or photographed documents; and certain Word formats (.docx) can also be processed.

What Happens Behind the Scenes

Understanding what Vibe does with your file helps explain the quality of its responses. Upon upload, it calls Mistral’s OCR 3 engine, which analyzes every page of the document, extracts the text, reconstructs the structure—headings, paragraphs, lists, tables—identifies images contained within the document and describes them, and then generates a faithful Markdown representation of the original. The result is clean, structured text that the language model then uses to answer your questions. All of this happens invisibly: you simply see the answer to your request, but it is this reconstruction step that allows the model to know a figure comes from a table or that a paragraph belongs to a specific section.

What to Ask a Document?

Once the document is uploaded, you can ask practically any question about its content. It helps to distinguish four categories. The first is factual: “What is the total amount on this invoice?”, “Who is the author of this report?”, “On what date was this contract signed?”, “What is the revenue figure mentioned on page 3?”. These are the most reliable questions because the answer is explicitly written in the document.

The second category is summarization. If you receive a 20-page report an hour before a meeting, prompts like “Summarize this document in 5 key points” or “Write an executive summary of this report” give you the essentials in seconds, while “What are the main arguments in this article?” prepares you for a debate. The third category is analysis, where the model cross-references and interprets: “What are the most restrictive clauses in this contract?”, “Compare the Q1 and Q2 results in this financial report”, or “Identify the risks mentioned in this document”. Finally, the fourth category is reformulation: “Rewrite section 3 in plain language”, “Translate the summary into English”, or—very useful for legal documents—“Explain this passage as if I were not a lawyer”.

Vibe handles multi-page documents without difficulty: the model has access to the entire content and navigates between pages to respond. You can even reference specific pages—“What does page 7 say about payment terms?” or “Compare the table on page 4 with the one on page 12”. For very long documents, over 50 pages, prefer targeted questions over a global summary: you will get significantly more precise answers.

Use Cases and Best Practices

Common use cases cover the entire professional daily routine: extracting amounts, dates, references, and terms from an invoice or quote; identifying key clauses, deadlines, and obligations in a contract; getting a summary of a report without reading all 30 pages; understanding the results and methodology of a scientific paper without the jargon; retrieving information filled in an administrative form; or analyzing a candidate’s skills and experience from their CV.

A few habits will significantly improve your results. File quality matters: a native PDF generated by software will yield better results than a poor-quality scan. Question precision matters too: “What is the pre-tax amount on page 2?” is more effective than a vague “Analyze this invoice”. To compare documents, upload them in separate messages and then ask for the comparison, rather than mixing everything together. For critical figures—financial amounts, contractual dates—always verify against the original document: extraction is reliable, but you remain responsible. Finally, keep in mind that the document is sent to Mistral’s servers for processing; avoid uploading extremely sensitive documents.

Key Takeaways

• Upload a PDF or document image to Vibe just like any other file
• The OCR 3 engine automatically extracts text, tables, and structure
• You can ask factual, summarization, analysis, or reformulation questions
• Multi-page documents are supported—reference specific pages in your questions
• Prefer precise questions to get the most reliable answers