Perplexity File Upload Limit: 2026 Caps That Bite

The Perplexity file upload limit is not one number: in 2026, a consumer researcher may need to keep practical uploads around 40 MB while Enterprise and Sonar workflows document a 50 MB ceiling, and that small difference can decide whether a research pack becomes searchable or stalls before analysis. i treat this as a planning problem, not a trivia question, because Perplexity now has several upload surfaces with different rules: Threads, Spaces, Personal Repository, Organisation Repository, File Connectors and the Sonar API.

For most readers, the quick answer is straightforward. Free and entry-level consumer use is best treated as limited and occasional, Pro is the sensible individual research plan, Enterprise Pro is the governance tier for teams, and Enterprise Max is the high-volume tier for persistent file libraries. The harder answer is that the visible file-size cap is only one constraint. File count, weekly upload allowance, retention period, connector support, prompt design and retrieval behaviour all shape whether Perplexity can analyse a file reliably.

During our 2026 evaluation, i compared Perplexity’s current help pages, enterprise file-limit documentation, pricing pages, Sonar API documentation and competing assistant limits from OpenAI, Anthropic, Google and Microsoft. The result is a practical map of what you can upload, where it lives, when it disappears, what it costs and when a larger plan is genuinely worth it. I also flag the limits that remain under-specified in public documentation, because professional teams should not build research or compliance workflows on guessed storage ceilings.

The Perplexity File Upload Limit in 2026

Perplexity’s upload rules make sense only after separating file size from file destination. The official Enterprise file-limit page documents Threads at up to 30 files per upload, each up to 50 MB, with retained Thread files deleted after 7 days. The same page describes Spaces, repositories and persistent totals separately, which means a file can have a different lifetime and governance profile even when the same 50 MB file-size ceiling applies.

This distinction matters for anyone using Perplexity as a research desk rather than as a casual chat box. A Thread upload is fast, ephemeral and ideal for a single question sequence. A Space is persistent, collaborative and context-rich. A repository is a personal or organisational source library. If you are new to the mechanics, the site’s guide on how to upload files to Perplexity AI remains a useful on-ramp, but the current enterprise documentation adds important caps for higher-volume teams.

The consumer picture is less perfectly public. Perplexity’s own plan pages describe Pro file uploads and Space capacity, while earlier consumer guidance and third-party testing frequently cite 40 MB as the working file-size limit for Free and Pro. For publishing purposes, I would not state a public 1 GB Perplexity upload cap as a verified general rule. I found no official public documentation that confirms a universal 1 GB upload allowance for the product surfaces covered here. Treat 40 MB as the conservative consumer margin and 50 MB as the documented Enterprise and Sonar ceiling unless your account console states otherwise.

Surface	Documented file size	Files per action	Retention	Best fit
Threads	Up to 50 MB in Enterprise docs	30 on Enterprise, lower on consumer plans	7 days for uploaded files	Fast, temporary question answering
Spaces	Up to 50 MB per file in Enterprise docs	500 files on Enterprise Pro, 5,000 on Enterprise Max	Persistent until deleted	Team projects and reusable research bases
Personal Repository	Up to 50 MB per file	5,000 files on Enterprise Pro, 10,000 on Enterprise Max	Persistent until deleted	Private working library
Organisation Repository	Up to 50 MB per file	500 files	Persistent until deleted	Approved shared sources
Sonar API media	50 MB per file	Up to 30 files per request where supported	Request and application dependent	Developer file analysis workflows

Why the Limit Changes by Surface, Not Just Plan

Perplexity’s product design is built around search-grounded answers, source memory and task context. That is why the same PDF can behave differently depending on whether it enters through a Thread, a Space or a connector. The citation-first search model also means file analysis is not isolated from retrieval. Uploaded files become part of a broader answer process in which Perplexity decides which source passages, web results and context blocks to use.

Threads are the lowest-friction route. They suit one-time analysis, such as asking for a contract summary, a table extraction or a comparison between two short reports. The trade-off is retention. If your team expects a file to remain available across future sessions, a Thread is the wrong storage surface.

Spaces move the file into a project context. That context can include instructions, collaborators and multiple documents. In practice, Spaces are better for recurring editorial, market research and academic workflows because they reduce the need to re-upload the same source pack. They also shift the problem from a one-file size cap to an information-architecture problem: naming, versioning and pruning files become part of prompt quality.

Repositories sit one layer deeper. A personal repository is designed for a user’s persistent knowledge base. An organisation repository gives administrators a way to make approved sources available. Enterprise buyers should care more about this layer than about one-off upload size, because repositories define what staff repeatedly ask the system to reason over. When an assistant repeatedly cites stale PDFs, the problem is rarely the model alone. It is usually a source-management failure.

Current Pricing Matrix and Hidden Caps

The commercial matrix shows why file limits are partly a pricing issue and partly a governance issue. Perplexity Pro is positioned for individual heavy users, while Enterprise Pro and Enterprise Max combine higher file volume with administrative controls. Official pricing lists Pro at $20 per month or $200 per year, Enterprise Pro at $40 per seat per month or $400 per seat per year, and Enterprise Max at $325 per seat per month or $3,250 per seat per year. Annual pricing is presented as a lower effective monthly rate.

The site’s own Pro versus free comparison is useful for consumer positioning, but enterprise buying decisions need the current Perplexity pricing page and Enterprise file-limit documentation beside it. The hidden commercial point is that a team buying Enterprise Pro for document research is not simply paying for more questions. It is paying for repository scale, Space capacity, data controls and administration.

Perplexity’s enterprise documentation also links volume to weekly behaviour. Enterprise Pro lists 100 Thread uploads per week, while Enterprise Max lists 1,000. That is a meaningful difference for analysts who batch filings, procurement documents or client research packs. Pro users with lighter research needs may never hit those weekly caps, while operations teams can reach them quickly during migration or discovery projects.

A second hidden cap is collaboration. The public pricing page describes limited collaborators on Pro Spaces and unlimited collaborators on Enterprise. Once a Space becomes a shared source room for a department, collaborator limits can matter as much as file count. A 50 MB cap is frustrating; a collaboration ceiling can stop a workflow entirely.

Plan or product	Published price signal	File or Space signal	Governance signal	Best buyer
Free or Standard	No paid monthly fee	Basic uploads, with small daily allowances noted in help docs	Minimal admin control	Light casual users
Pro	$20 monthly or $200 yearly	Pro uploads and up to 50 files per Space on current plan descriptions	Individual controls and opt-out options	Solo researchers and writers
Max	$200 monthly consumer plan signal in 2026 coverage	Higher usage credits than Pro, not a team storage substitute	Individual premium use	Power users testing advanced features
Enterprise Pro	$40 per seat monthly or $400 yearly	100 weekly Thread uploads, 500 Space files, 15,000 persistent files	SSO, admin, data protection, organisation repository	Teams with controlled knowledge bases
Enterprise Max	$325 per seat monthly or $3,250 yearly	1,000 weekly Thread uploads, 5,000 Space files, 50,000 persistent files	Largest governed file volumes	Departments with sustained document operations
Sonar API	Usage based API pricing separate from app plans	50 MB per file and up to 30 files per request where supported	Developer controlled retention and app policy	Builders integrating Perplexity into products

Threads: Fastest Upload Path, Shortest Retention

Threads remain the simplest place to upload a file because the user does not have to design a knowledge base first. Attach the file, ask a specific question and refine the answer. That is why Threads are still the best path for one-off summaries, short due diligence, meeting-note analysis and quick spreadsheet interpretation. They are also where the shortest retention rule bites hardest: uploaded Thread files are retained for 7 days under Perplexity’s data collection guidance and Enterprise file-limit documentation.

Speed is the advantage. Users who rely on browser shortcuts can accelerate repeated research sessions by pairing upload behaviour with the site’s keyboard shortcut workflow. In practice, the fastest Thread workflow is not to upload every source and ask for a general answer. It is to upload a narrow pack, ask for a defined extraction, then ask Perplexity to cite where the answer came from inside the uploaded material.

The documented Enterprise Thread limit is up to 30 files per upload, up to 50 MB per file. Perplexity also documents one practical limitation that is easy to miss: a single query supports one upload method and one online storage service at a time. In other words, mixing local files, Google Drive and another storage source in one query can break the mental model. Teams should standardise how they attach files to a Thread rather than relying on ad hoc combinations.

Thread uploads should be treated as temporary scratch material. If a file will be cited in a report, reused in a future briefing or shared with colleagues, move the source into a Space or repository. If the file contains confidential material, the decision should also pass through privacy settings and enterprise data policy. The upload button is not a retention policy.

Thread workflow for cleaner answers

Use one project topic per Thread, keep the file pack narrow, ask for page-level evidence and save only the final verified source set into a persistent surface. This reduces hallucinated cross-references and makes later audit easier.

Spaces and Repositories: Persistent Files, Bigger Planning Problem

Spaces are the point where Perplexity starts to feel less like a chatbot and more like a shared research room. The Perplexity Spaces guide is useful because it frames Spaces as persistent contexts rather than isolated prompts. Current Enterprise file-limit documentation puts Enterprise Pro Spaces at 500 files and Enterprise Max Spaces at 5,000 files, each up to 50 MB. Pro plan pages and site-level guidance often describe 50 files per Space for individual use, which is enough for many editorial and research workflows.

Repositories extend persistence beyond one Space. Perplexity’s Enterprise file-limit documentation lists 5,000 personal repository files for Enterprise Pro and 10,000 for Enterprise Max. It also lists persistent combined totals of 15,000 files for Enterprise Pro and 50,000 for Enterprise Max across repositories and Spaces. These figures change the evaluation from upload convenience to lifecycle management.

A persistent source base should be curated. If analysts upload every draft, duplicate PDF and outdated board deck, Perplexity can still answer, but retrieval becomes noisier. The best enterprise deployments use a simple source taxonomy: canonical documents in the organisation repository, project-specific materials in Spaces, personal drafts in the personal repository and disposable analysis files in Threads.

The biggest planning mistake is to buy a larger tier before fixing source hygiene. Enterprise Max can hold vastly more files, but it cannot make redundant source packs clean by magic. Volume only helps when the repository structure, naming conventions and file-retention rules are clear before staff begin uploading at scale.

Persistent surface	Enterprise Pro	Enterprise Max	Operational note
Spaces	500 files per Space	5,000 files per Space	Best for project or team contexts
Personal Repository	5,000 files	10,000 files	Private user source library
Organisation Repository	500 files	500 files	Approved organisation-wide source set
Combined persistent files	15,000 files	50,000 files	Repository plus Spaces total per user
Weekly Thread uploads	100 per week	1,000 per week	Important during migration and batch analysis

API and Sonar Media: Building With File Inputs

For developers, Perplexity’s Sonar media documentation is the clearest public file-input specification. It documents a 50 MB maximum per file, supports PDF, DOC, DOCX, TXT and RTF file types, and describes both file URL and base64 input patterns. It also states that the base64 payload must be only the base64 string, without a data URL prefix. That small implementation detail is exactly the kind of thing that breaks production prototypes.

The API documentation also states that requests can include multiple files, up to 30 where supported. The application consequence is straightforward: a developer can build a document-answering workflow around Perplexity, but must still design for file packaging, retries, validation and source traceability. The API is not a bulk storage system. It is a request-time analysis path that should sit beside your own document store, audit logs and deletion rules.

Perplexity’s 2026 changelog also points to a broader developer strategy. The company has discussed custom remote MCP connectors, prebuilt connectors and API surfaces including Agent API, Search API, Embeddings API and Sandbox API. In the same period, Perplexity framed Computer as an orchestrated system for files, tools, memory, models and the internet. Aravind Srinivas, Perplexity’s CEO, wrote that “Files, tools, memory, and models” are orchestrated together. That is a concise statement of the direction of travel: file uploads are becoming inputs to task systems, not just attachments to chats.

The technical bottleneck is therefore orchestration, not simply upload size. Production teams should validate MIME type, file size, token density, page count, scanned text quality and expected retrieval output before submitting to Sonar. A file that is under 50 MB can still be a poor input if it contains low-quality scans, large embedded images or tables with inconsistent structure.

Supported Formats, Multimodal Analysis and Connector Gaps

Perplexity’s file support is broad enough for most professional research, but not uniform across surfaces. Official File Connector documentation for Enterprise lists Google Docs, Google Slides and Google Sheets; Microsoft Office formats including DOCX, XLSX, PPTX and XLS; PDF; CSV; RTF; ODT; Markdown; JSON; and TXT. It also states that connectors do not support images, audio or video, although those formats can be uploaded directly to Threads. This is where the site’s discussion of the best Perplexity features needs a current caveat: the feature is powerful, but the upload surface decides the file-type pathway.

For document-heavy research, the supported set is sensible. PDFs, Word documents, spreadsheets, CSV exports and plain text cover contracts, papers, meeting notes, financial models and policy documents. For media-heavy research, direct Thread upload may be more appropriate than connectors, but users should check current in-product allowances because image and media handling can vary by plan and region.

The most common failure mode is not unsupported format, but unsupported expectations. A scanned PDF may appear to upload while still producing weak answers if OCR is poor. A spreadsheet may pass the file-size cap while containing hidden tabs, formulas or merged cells that the model does not interpret as the user expects. A JSON file may be syntactically valid but too deeply nested for convenient answer extraction.

During our 2026 evaluation, i found that the safest pre-upload format choice is the simplest faithful representation of the source. Convert scanned files to searchable PDFs. Export analysis tables to clean CSV when formulas are not needed. Keep raw JSON separate from a short schema explainer. For long reports, include an executive summary file that tells Perplexity what the document pack contains before asking for cross-document synthesis.

How to Upload, Structure and Test Large Files

A practical upload workflow starts before Perplexity sees the file. First, classify the task. If the job is a one-off summary, use a Thread. If it is a reusable project, use a Space. If it is a canonical source, use a repository. Second, reduce the file to the smallest faithful version. Remove decorative images, compress PDFs, export oversized spreadsheets to CSV and split giant reports by chapter or appendix.

Third, create a manifest. For any pack with more than five files, add a short text file listing file names, dates, source owners and the question each file is supposed to answer. This is not glamorous, but it gives Perplexity a map and gives the human reviewer a way to spot missing context. Fourth, ask a verification question before asking the final research question. For example: identify every file in the pack, list its apparent topic and cite one page or section from each file. If Perplexity misses a file, the final synthesis is not ready.

Fifth, test retrieval with controlled questions. Ask for a figure that appears only in one document, a definition that appears in a specific appendix and a contradiction between two sources. This reveals whether the uploaded files are actually being used or whether the model is leaning on web context. Sixth, keep an audit trail outside Perplexity. Save the original files, prompts, answer versions and source checks in your document-management system.

This workflow is slower than drag-and-drop, but it prevents expensive rework. The limit that matters in professional research is not just whether a 49 MB file uploads. It is whether the uploaded material can be found, cited, challenged and reused without someone rebuilding the context from scratch next week.

Perplexity file upload limit checklist

Keep consumer files below the safest published working margin where possible, use 50 MB as the documented Enterprise and Sonar ceiling, avoid mixing upload methods in one query, split long PDFs by section, convert scanned pages with OCR, add a manifest and test source recall before relying on an answer.

Hands-On Evaluation: Where Bottlenecks Appear

When we integrated this evaluation into an editorial research workflow, the bottleneck was rarely a single file-size number. It was the combination of page length, source structure and prompt scope. Academic users see this quickly, which is why the site’s academic research workflow is a natural companion to file-limit planning. A 30 MB paper with dense tables can be harder to interrogate than a 48 MB slide deck with clear headings.

During our 2026 evaluation, i used reproducible desk testing rather than confidential enterprise tenant testing. The sample packs included short PDFs, long reports, CSV exports, policy documents and mixed file bundles mapped against Perplexity’s public limits. The aim was not to claim private account behaviour. It was to identify where publicly documented caps turn into workflow friction.

The first bottleneck was selective extraction. Site-level guidance and hands-on document workflows indicate that long files may be selectively extracted rather than read in a perfectly exhaustive way. That is normal for retrieval systems, but it changes prompt design. Instead of asking, summarise this 180-page report, better prompts ask for page-specific findings, section-by-section risk registers or a table keyed to exact headings.

The second bottleneck was cross-file synthesis. As the number of files rises, Perplexity must decide which passages matter. A manifest improves that routing. The third bottleneck was spreadsheet interpretation. Large spreadsheets need pre-cleaned tabs and explicit instructions about formulas, hidden sheets and units. The fourth bottleneck was retention. A result that depends on Thread uploads should be saved, cited and migrated quickly if it will support published work or client advice.

The distinctive insight is simple: upload limits describe the front door, while retrieval quality describes the room inside. Professionals need both.

Perplexity vs ChatGPT, Claude, Gemini and Copilot

Perplexity is not the largest-file assistant by raw megabyte ceiling. OpenAI’s File Uploads FAQ describes a 512 MB hard limit for most uploaded files, a 2 million token limit for documents, approximate 50 MB caps for spreadsheets and larger account-level storage caps. Anthropic’s 2026 support documentation lists Claude chat uploads up to 500 MB per file and up to 20 files per chat, while Claude Projects use smaller project-file limits. Google’s Gemini Apps documentation lists 10 files in the same prompt, 100 MB for many non-video files and 2 GB video inputs, while the Gemini API Files service documents 2 GB per file and 20 GB per project storage with 48-hour storage. Microsoft Copilot’s consumer file-upload support lists up to 20 files in one conversation and 50 MB per file.

This makes Perplexity’s value proposition more specific. It is strongest when a user wants search-grounded synthesis, source citations, Spaces and managed research context. It is weaker if the only requirement is the largest possible single upload. Lucia Loher, Product Manager for Gemini API, captured the competitive direction when Google announced it had “increased the file size limit from 20MB to 100MB.” The market is pushing file inputs upward, but larger caps do not automatically produce better answers.

Dmitry Shevelenko, Perplexity’s Chief Business Officer, framed Computer for Enterprise as the company’s “single biggest productivity unlock.” That matters because file uploads are becoming part of enterprise agent workflows. If agents can plan, browse, retrieve, write and use tools, the file limit is one of several rails that determine whether a workflow reaches production quality.

Assistant	Single-file signal	File count signal	Storage or retention signal	Practical reading
Perplexity	50 MB documented for Enterprise and Sonar, 40 MB safer consumer margin	30 Enterprise Thread files, 500 or 5,000 Space files by tier	7-day Thread retention, persistent Spaces and repositories	Best for citation-led research contexts
ChatGPT	512 MB hard file cap, 2 million tokens for documents	80 files every 3 hours for paid tiers, 3 daily for Free	25 GB user cap and 100 GB org cap	Strong for large files and broad analysis
Claude	Up to 500 MB per chat file in 2026 support docs	Up to 20 files per chat	Project files have separate smaller limits	Strong for large conversational file packs
Gemini	100 MB many app files, 2 GB video, 2 GB per API file	10 files per app prompt, API project storage model	48 hours in Gemini API Files service	Strong for media and developer pipelines
Microsoft Copilot	50 MB in consumer file upload support	Up to 20 files per conversation	Conversation scoped	Convenient for Microsoft ecosystem users

Enterprise Governance, Connectors and Data Retention

Enterprise buyers should evaluate Perplexity’s file limits alongside privacy and retention. Perplexity’s data collection help page states that Pro and Max subscribers can opt out of data use for training, while Enterprise data is not used for model training. It also states that files uploaded to Threads are retained for 7 days, with additional custom retention options for eligible administrators. For regulated teams, that retention language is as important as the 50 MB size limit.

The scale question is also a market question. Perplexity has grown from a consumer answer engine into a business research platform, and the site’s verified Perplexity user metrics show why enterprises now care about operational controls rather than just clever answers. More usage means more source material, more accidental uploads and more need for policy.

Connectors add another governance layer. Perplexity’s Enterprise File Connector documentation supports Google and Microsoft file types, PDFs, CSVs, Markdown, JSON and text formats. It allows files and folders to be synced, but it does not support adding entire drives. It also does not support images, audio or video through connectors. Those constraints are sensible security boundaries, but administrators need to communicate them clearly to users.

Aravind Srinivas described Perplexity Computer as bringing “the file system, tools” into one system. The phrase is short, but it captures the governance challenge. Once files are not just passive uploads but active inputs to agents, every organisation needs rules about what can be connected, what can be retained, who can share it, how stale sources are retired and how answers are audited. Without that layer, a larger upload allowance can simply create a larger compliance problem.

Which Plan Should You Choose?

Choose the smallest plan that matches your file lifecycle, not the largest plan that sounds safe. For students, independent writers and solo analysts, Pro is usually the practical starting point because it expands research capacity without forcing an enterprise administration model. It suits PDFs, smaller source packs, recurring Spaces and day-to-day web-grounded research.

Choose Enterprise Pro when files are shared by a team, when source access needs to be controlled, or when the organisation wants persistent repositories, SSO and stronger administrative controls. The 100 weekly Thread uploads and 15,000 persistent file ceiling are sufficient for many research, legal, product and strategy teams if the source library is curated.

Choose Enterprise Max only when the storage and upload multiplier is real, not hypothetical. Its 1,000 weekly Thread uploads, 5,000 files per Space and 50,000 persistent files per user are meaningful for teams handling large evidence bases, high-frequency research operations or document-heavy agent workflows. If the team lacks naming conventions and retention policy, Max can amplify clutter.

Choose Sonar API when you need Perplexity inside a product, internal research tool or automated pipeline. The API route should include validation, logging, retries and human review for high-stakes outputs. It should not be treated as a replacement for document storage. The safest rule is this: Pro buys better individual research, Enterprise buys governance, Max buys volume and Sonar buys integration.

Takeaways

• Treat the Perplexity file upload limit as a surface-specific rule: Threads, Spaces, repositories and Sonar do not behave the same way.
• Keep consumer uploads below the safer 40 MB working margin when possible, unless your account interface or current Perplexity documentation confirms a higher allowance.
• Use 50 MB as the documented ceiling for Enterprise file surfaces and Sonar media inputs in current public documentation.
• Do not rely on Thread uploads for lasting evidence. Thread files are documented as retained for 7 days, so save and migrate important sources quickly.
• Use Spaces for recurring research, repositories for canonical sources and Threads for disposable analysis.
• Split long PDFs, convert scans with OCR, add a manifest and ask source-verification prompts before trusting cross-file synthesis.
• Buy Enterprise Pro for governance and team file persistence; buy Enterprise Max only when high-volume file work is already disciplined.
• For developers, treat Sonar file upload as an analysis input and keep storage, audit, deletion and retry logic in your own application stack.

Our Content Testing Methodology

This guide was compiled through a 2026 feature-limit verification process focused on Perplexity’s live plan pages, Enterprise file-limit documentation, Enterprise pricing page, Sonar media documentation, File Connector help, data collection policy and account-management guidance. I cross-checked those sources against current upload documentation from OpenAI, Anthropic, Google Gemini and Microsoft Copilot, then mapped sample file packs against the documented limits for file size, file count, persistence, connector support and retention. The desk tests used clean PDFs, long reports, CSV exports and mixed document bundles to identify practical bottlenecks such as selective extraction, cross-file routing and spreadsheet ambiguity. I did not test confidential Enterprise tenant-only behaviour or unpublished 1 GB contexts, so those claims are deliberately not presented as verified public limits.

Conclusion

Perplexity’s file limits now sit at the intersection of search, storage and enterprise workflow design. The headline number is useful, but it is incomplete. A 50 MB documented cap for Enterprise and Sonar inputs tells only part of the story; the real operating question is where the file lives, who can use it, how long it remains available and whether Perplexity can retrieve the right passage when the answer matters.

The direction is clear. Perplexity is moving from answer engine to work system, with Spaces, repositories, connectors, Sonar and Computer-style orchestration pulling files into broader tasks. That makes file governance more important, not less. Larger caps will probably arrive across the industry, as Google, OpenAI, Anthropic and Microsoft keep expanding file input surfaces. The open question is whether larger uploads will be matched by better source controls, clearer retention settings and more transparent retrieval behaviour. For now, the best strategy is disciplined: keep files clean, choose the right surface, verify source recall and avoid building workflows on unpublished limits.

FAQs

What is the Perplexity file upload limit in 2026?

The safest public answer is that Perplexity’s limit depends on surface and plan. Enterprise and Sonar documentation show 50 MB per file in covered contexts. Consumer guidance is less explicit, so 40 MB remains the safer working margin for Free and Pro unless the live product interface confirms otherwise.

How many files can I upload to a Perplexity Thread?

Perplexity’s Enterprise documentation lists up to 30 files per Thread upload, with each file up to 50 MB. Consumer Thread allowances are lower and plan dependent. Thread uploads are best for temporary analysis because uploaded Thread files are documented as retained for 7 days.

How many files can a Perplexity Space hold?

Enterprise Pro Spaces are documented at 500 files per Space, while Enterprise Max Spaces are documented at 5,000 files per Space. Pro plan guidance commonly describes 50 files per Space for individual use. Exact availability can vary by current plan page and account interface.

Does Perplexity analyse images or only text documents?

Perplexity can handle more than plain text, but support depends on upload path. Enterprise File Connectors support documents, spreadsheets, PDFs, CSV, Markdown, JSON and text formats, but not images, audio or video. Those media types may be available through direct Thread uploads instead.

How long are files stored in Perplexity?

Perplexity’s data guidance says files uploaded to Threads are retained for 7 days. Files in Spaces and repositories are persistent until deleted, subject to plan limits and administrator policy. Enterprise administrators may have additional retention controls in eligible deployments.

What file formats does Perplexity support?

Perplexity’s Sonar media documentation lists PDF, DOC, DOCX, TXT and RTF. Enterprise File Connectors add Google Docs, Slides, Sheets, Microsoft Office formats, PDFs, CSV, RTF, ODT, Markdown, JSON and TXT. Support still depends on upload surface.

Is Perplexity Pro enough for document research?

Yes, for many solo researchers. Pro is suitable for routine PDFs, small research packs and Spaces. It is not the best fit when a team needs persistent shared repositories, enterprise retention policy, SSO, administrator controls or high weekly upload volume.

Does the Sonar API support file uploads?

Yes. Perplexity’s Sonar media documentation supports file inputs, including URL and base64 patterns, with a 50 MB maximum per file and up to 30 files per request where supported. Developers should validate file type, file size and source logging in their own systems.

References

1. Perplexity. (2026a). Enterprise file limits. Perplexity Help Center. https://www.perplexity.ai/help-center/en/articles/12137006-enterprise-file-limits
2. Perplexity. (2026b). Enterprise pricing. Perplexity. https://www.perplexity.ai/enterprise/pricing
3. Perplexity. (2026c). Media and attachments. Perplexity API Docs. https://docs.perplexity.ai/guides/media-uploads
4. Perplexity. (2026d). Introduction to file connectors for Enterprise organizations. Perplexity Help Center. https://www.perplexity.ai/help-center/en/articles/11921327-introduction-to-file-connectors-for-enterprise-organizations
5. Perplexity. (2026e). Data collection at Perplexity. Perplexity Help Center. https://www.perplexity.ai/help-center/en/articles/10352901-data-collection-at-perplexity
6. OpenAI. (2026). File uploads FAQ. OpenAI Help Center. https://help.openai.com/en/articles/8555545-file-uploads-faq
7. Anthropic. (2026). Upload files to Claude. Anthropic Support. https://support.anthropic.com/en/articles/8241126-upload-files-to-claude
8. Google AI for Developers. (2026). Files API. Gemini API Documentation. https://ai.google.dev/gemini-api/docs/files
9. Nuñez, M. (2026, March 10). Perplexity takes its Computer AI agent into the enterprise, taking aim at Microsoft and Salesforce. VentureBeat. https://venturebeat.com/technology/perplexity-takes-its-computer-ai-agent-into-the-enterprise-taking-aim-at-microsoft-and-salesforce/