top of page

Perplexity AI file upload support: limits, formats, and usage in 2025

ree

Perplexity AI’s Sonar platform has steadily expanded its file-handling capabilities, enabling richer in-context analysis for research, coding, and collaborative work. As of September 2025, users can upload a wide range of file types—documents, spreadsheets, slide decks, code snippets, images, audio, and even short video—to receive grounded, citation-ready answers powered by Sonar’s retrieval engines. This guide summarizes the latest size limits, format coverage, and workspace behaviors across Free, Pro, and Enterprise plans.



Standard uploads cover most common formats and media types.

Perplexity supports the following extensions: PDF, DOCX, PPTX, XLSX, CSV, RTF, ODT, MD, JSON, TXT for documents and code; PNG, JPEG, WEBP, GIF for images; and MP3, WAV, MP4 for audio/video. Video uploads focus on transcript extraction and key-frame analysis—scene-level search is still in development.


File size limits increase with each subscription tier.

Tier

Max file size

Files per prompt

Retention window

Free / Plus

40 MB

10

30 days

Pro

50 MB

10

90 days

Enterprise Pro

1 GB

20

1 year (customizable)

For API users, base-64 image payloads may reach 50 MB, and external image URLs bypass the size cap.



Spaces allow organizing larger corpora for team collaboration.

Perplexity’s Spaces feature functions like project folders. Pro subscribers can store up to 50 files per Space, while Enterprise customers can negotiate higher quotas. All members of a Space can reference its files in prompts without re-uploading, preserving context across shared research threads.


File Connectors synchronize cloud documents automatically.

Enterprise Pro adds File Connectors that pull content from Google Drive, Docs, Sheets, and Slides on a schedule, indexing material for internal search and Sonar prompts. Connectors respect existing permissions, and media files (images, audio, video) still require manual upload to ensure intentional ingestion.


The internal index supports hundreds of files for enterprise knowledge bases.

By default, an Enterprise tenant can index 500 files—spanning documents, spreadsheets, and slide decks—for private semantic search. Administrators can request larger allocations or set exclusion rules based on path or MIME type.


Practical tips help manage quotas and optimize analysis quality.

  • Compress PDFs or split large decks to stay under tier limits.

  • Consolidate code snippets into a single Markdown file to reduce prompt file counts.

  • For video, ensure clear audio tracks—Sonar currently prioritizes speech transcription over scene recognition.

  • Use Spaces to stage research corpora, then reference them with concise prompts rather than re-uploading.

  • Enterprise admins should enable audit logging on File Connectors to track sync activity and compliance.



Perplexity AI’s file-handling roadmap focuses on deeper multimedia understanding.

Upcoming quarterly releases aim to deliver:

  • Scene-level video search with key-frame extraction.

  • Optical character recognition on scanned images and PDFs.

  • Inline code execution for uploaded notebooks in Pro.

These enhancements will further embed Perplexity into complex research and data-driven workflows, making file uploads an even more integral part of the Sonar experience.


____________

FOLLOW US FOR MORE.


DATA STUDIOS


bottom of page