# Perplexity Usage Limits: Free, Pro, and Enterprise Search and File Upload Caps

Perplexity usage limits define the allowances for standard queries, multi-step Pro searches, session upload caps, and persistent Project boundaries across account tiers. Free accounts receive 5 Pro searches every 4 hours and 3 file uploads daily, while Pro subscribers receive 300+ daily Pro queries. When research archives exceed native session attachment caps or persistent Project boundaries, intelligent workspaces provide external indexing through the Model Context Protocol.

Source: https://fast.io/resources/perplexity-usage-limits/
Author: [Tom Langridge](https://fast.io/authors/tom-langridge/)
Last reviewed: 2026-10-28

## What Are Perplexity Usage Limits Across Account Tiers?

Perplexity caps free accounts at 5 Pro searches every 4 hours and 3 file uploads daily, while Pro subscribers receive 300+ daily Pro queries and standard searches remain unmetered across all plans. Perplexity usage limits define the volume of standard searches, Pro searches (typically 300+ daily queries on Pro compared with 5 every 4 hours on Free accounts), and file uploads (typically 500+ on Pro compared with 3 to 4 daily on Free accounts) available per account tier. For researchers, data analysts, and engineering teams, having perplexity search limits explained across tiers prevents sudden interruptions during critical investigation workflows.

Perplexity separates conversational retrieval into distinct operational query modes. The mode an account invokes determines execution paths, backend model routing, and quota drawdown:

* **Standard Search:** The default search experience executes single-pass web retrieval using Perplexity optimized base models. Standard search answers factual questions, summarizes linked sources, and generates quick answers without drawing down premium quota allowances. Standard search remains unmetered across all plans, including Free accounts.

* **Pro Search:** An advanced multi-step research workflow that analyzes complex prompts, executes dynamic follow-up web crawls, evaluates dozens of primary sources, executes Python code for data analysis, and synthesizes findings with direct citations. Pro Search supports model switching across frontier reasoning systems, including Claude Sonnet, GPT-4o, and Sonar reasoning models.

* **Research Mode (Deep Research):** An extended analytical mode designed for comprehensive literature reviews and exhaustive market research, generating multi-page structured reports.

### The Mechanics of Rolling Usage Quotas

Perplexity manages Pro query allowances through rolling usage windows rather than fixed calendar resets. Under this framework, an account quota refreshes continuously rather than resetting on a fixed calendar schedule. When a subscriber submits a Pro query, that execution remains recorded in the active usage budget for a rolling window. As earlier queries age past the rolling threshold, capacity becomes available automatically. This architecture provides smoother workload distribution, though it makes tracking exact remaining perplexity query limits less obvious from the interface.

### Structured Plan Comparison: Search and Query Allocations

The table below details documented query limits, model availability, and reset structures across consumer and organizational tiers, checked as of October 2026:

| Account Tier | Monthly Cost | Pro Searches | Daily File Uploads | Max File Size | Reset Schedule |
| :--- | :--- | :--- | :--- | :--- | :--- |
| **Free (Standard)** | $0 | 5 queries per 4 hours | 3 files daily | 25 MB to 40 MB | Fixed 4-hour rolling cycle |
| **Pro** | $20 | 300+ queries per day | 50+ files daily (rolling) | 40 MB (100 MB in Projects) | Continuous rolling 24-hour window |
| **Enterprise Pro** | Custom | 400 queries per week | 30 files per upload, 100 per week | 50 MB (100 MB in Projects) | Contractual weekly allocation |
| **Enterprise Max** | Custom | 4,000 queries per week | 30 files per upload, 1,000 per week | 50 MB (100 MB in Projects) | Contractual weekly allocation |

### Model Switching and High-Demand Fallback Behavior

When high-demand foundation models experience sudden regional traffic surges or capacity limits, Perplexity maintains conversational continuity through automated fallbacks. If an account exhausts its included Pro Search allowance or if an upstream model cluster becomes overloaded, the platform redirects queries to unmetered Standard search mode. This ensures basic search functionality never shuts down entirely, although multi-turn reasoning and deep source synthesis pause until rolling allowances clear.

## What Are the File Upload Caps for Sessions and Collections?

Beyond text-based queries, Perplexity allows users to upload local files directly into search sessions to provide immediate analytical context. Document ingestion introduces specific capacity constraints that differ between session-level attachments and long-term project storage. Most user guides mix up standard queries with Pro searches and omit the 25 MB per-file upload ceiling that Perplexity enforces on file attachments in mobile sessions and Collections.

According to official [Perplexity Help Center](https://www.perplexity.ai/help-center/en/articles/10354807-file-uploads) documentation, Perplexity caps individual file attachments at 40 MB across all supported document and media formats in standard search sessions. When an individual file attachment in a standard search session exceeds this 40 MB ceiling, Perplexity halts processing and returns an upload error.

In practice, users encounter two distinct file size boundaries depending on their interface. While the web interface supports 40 MB per file in active search sessions, mobile applications, Collection attachments, and certain file types encounter a stricter 25 MB per-file upload ceiling. Understanding this secondary boundary prevents unexpected upload rejections when attaching large scanned legal documents, complex financial models, or dense research PDFs.

The table below outlines how file upload and storage parameters vary across different operational areas:

| Storage Context | Account Tier Eligibility | Maximum File Size | Upload Volume Allowance | Retention Schedule |
| :--- | :--- | :--- | :--- | :--- |
| **Search Session Attachments** | Free (Standard) | 25 MB to 40 MB | 3 files daily | Retained 7 days, then deleted |
| **Search Session Attachments** | Pro | 40 MB per file | 50+ files daily (rolling weekly) | Retained 7 days, then deleted |
| **Enterprise Ask Attachments** | Enterprise Pro | 50 MB per file | 30 files per upload, 100 per week | Retained 7 days, then deleted |
| **Enterprise Ask Attachments** | Enterprise Max | 50 MB per file | 30 files per upload, 1,000 per week | Retained 7 days, then deleted |
| **Projects (Persistent)** | All tiers | 100 MB per file | 10,000 files and 5 GB total capacity | Retained until manually deleted |

### The Perplexity File Limit per Day Across Account Tiers

Attaching files to a search session does not grant an unlimited upload stream. Document allowances are governed by account-level rate limits:

* **Free Tier Allowances:** The perplexity free tier limits file ingestion to basic exploratory use, restricting accounts to 3 file uploads daily before prompting an upgrade to Pro.

* **Pro Tier Allowances:** Paid Pro subscribers receive expanded file ingestion quotas that accommodate 50+ file uploads per day, typically evaluated across a rolling weekly allowance of 500 or more files. Rather than enforcing a rigid perplexity file limit per day that cuts off work unexpectedly, Pro accommodates regular analytical workflows while monitoring high-volume automated abuse.

* **Enterprise Session Caps:** Enterprise Pro accounts can upload 30 files per query session, with a cumulative ceiling of 100 files per week across the team. Enterprise Max expands this ceiling to 1,000 files per week across sessions.

* **Query Concurrency:** Standard consumer sessions support attaching 10 files in a single prompt. Image uploads follow similar rules, with 30 images accepted per query.

### Audio and Video Processing Capabilities

Perplexity supports media file uploads alongside text documents, accepting formats such as MP3, WAV, FLAC, AAC, MP4, MOV, AVI, and WEBM. When media files enter a session, the system applies specialized parsing pipelines:

1. **Automated Speech-to-Text:** The ingestion engine transcribes spoken audio into structured textual transcripts.

2. **Speaker Identification:** The transcription pipeline diarizes conversations, tagging individual speakers throughout the recording.

3. **Natural Language Search:** The resulting transcript becomes searchable by the model, allowing users to ask targeted questions about spoken meetings, lectures, and interviews.

A critical technical boundary exists for video files: Perplexity does not index visual scenes, onscreen graphics, or frame sequences. Only the extracted audio transcript is processed. Users attempting to analyze video slides, software demonstrations, or visual diagrams will find that the model cannot reference visual contents that lack spoken descriptions.

### Ephemeral Retention and Quota Recovery Windows

Session file uploads are designed for temporary conversational analysis rather than permanent storage. Files attached directly to search sessions are retained for 7 days, after which they are permanently deleted from Perplexity storage caches.

When an account reaches its upload quota, the interface presents an upgrade notification. This message indicates that your rolling window is full; it does not mean file upload features have been removed from your plan. Because capacity is evaluated on a continuous rolling timeline, individual uploads roll out of the calculation approximately 168 hours after submission. Capacity recovers incrementally throughout the week rather than resetting all at once.

## What Storage Limits Apply to Perplexity Projects and Cloud Connectors?

Because session-level attachments expire after 7 days, users managing ongoing research initiatives rely on Perplexity Projects. Projects provide a persistent workspace where uploaded reference files, system instructions, and historical threads remain accessible across multiple sessions.

Despite their persistent nature, Projects operate under strict universal storage ceilings. These ceilings apply identically across all subscription levels, meaning an Enterprise workspace faces the exact same individual Project constraints as a standard Pro account.

### Universal Project Boundaries: File Counts and Repository Ceilings

Perplexity establishes strict structural boundaries on every Project to balance backend retrieval load:

* **File Count Ceiling:** A single Project holds a maximum collection of ten thousand documents.

* **Total Storage Ceiling:** As shown in the storage summary table above, cumulative storage across all documents within one Project operates under a strict repository ceiling.

* **Per-File Size Ceiling:** Individual files added to a Project can reach larger dimensions than session attachments, but single-document ingestion thresholds still block massive dataset archives.

Once a Project reaches either its total file count or storage capacity boundary, subsequent file uploads and third-party connector imports are immediately rejected. Teams managing extensive reference libraries must either distribute documents across multiple separate Projects or curate their collections to stay beneath aggregate Project boundaries.

### Cloud Connectors and the Static Snapshot Trap

To simplify document ingestion, Perplexity provides File Connectors that link directly to external cloud storage platforms, including Google Drive and Dropbox. While these connectors simplify file selection, engineering teams must recognize a fundamental architectural behavior:

Files imported through File Connectors are one-time static snapshots. Perplexity does not maintain a continuous synchronization pipeline with your cloud storage provider.

If a team member edits a technical specification, updates a pricing spreadsheet, or deletes an obsolete contract in Google Drive, the corresponding copy stored in your Perplexity Project remains unchanged. The AI model continues to answer questions based on the historical version captured at the moment of import. To work from current information, users must manually re-fetch and re-index the updated files. Relying on File Connectors for dynamic, rapidly evolving codebases or living policy manuals introduces significant risk of hallucination based on stale data.

### Enterprise Document Management and Scoped Projects

For enterprise organizations, document sharing is scoped to specific Projects and File Connectors rather than an unsegmented global repository. Enterprise administrators and team members organize reference documents within dedicated Projects or link to cloud storage through File Connectors, keeping collaboration bounded by team and project contexts.

## What Rate Limits Apply to the Perplexity Search and Agent APIs?

Developers and automation engineers accessing Perplexity programmatically interact with an entirely different rate-limiting framework. While consumer and team plans meter queries by rolling weekly allowances, the Perplexity API evaluates throughput using cumulative financial spending tiers and a leaky bucket rate-limiting algorithm.

API accounts advance through progressive usage tiers based on the total volume of credits purchased over the lifetime of the account. Cumulative spend reflects total lifetime purchases rather than current prepaid credit balances. According to official [Perplexity API Documentation](https://docs.perplexity.ai/docs/admin/rate-limits-usage-tiers), higher tiers unlock higher request and token limits, and once you reach a tier, you keep it permanently with no downgrade.

### Cumulative Spending Tiers for the Agent API

The table below outlines cumulative spend thresholds, Agent API request velocity, and Embeddings API limits across platform tiers, checked as of October 2026:

| API Tier | Lifetime Credit Purchases | Agent API Request Limit | Requests Per Minute (RPM) | Embeddings API Limit | Target Operational Profile |
| :--- | :--- | :--- | :--- | :--- | :--- |
| **Tier 0** | $0 | 1 QPS | 50 RPM | 85 QPS | New developer accounts and sandbox testing |
| **Tier 1** | $50+ | 3 QPS | 150 RPM | 170 QPS | Early prototyping and light internal scripts |
| **Tier 2** | $250+ | 8 QPS | 500 RPM | 170 QPS | Regular application production |
| **Tier 3** | $500+ | 17 QPS | 1,000 RPM | 170 QPS | Heavy production workloads |
| **Tier 4** | $1,000+ | 33 QPS | 4,000 RPM | 335 QPS | Large-scale production and high concurrency |
| **Tier 5** | $5,000+ | 33 QPS | 8,000 RPM | 335 QPS | Enterprise-scale automated infrastructure |

### Search API Leaky Bucket Mechanics

A critical distinction in the Perplexity developer ecosystem exists between general model endpoints and the dedicated Search API (`POST /search`). Unlike agent endpoints that scale with spending tiers, the Search API enforces a universal rate limit across all accounts regardless of tier status:

* **Sustained Throughput:** 50 query units per second.

* **Burst Capacity:** 50 query units instantly.

* **Unit Consumption:** A single query consumes 1 query unit. A multi-query request array consumes 1 query unit for each individual query string in the payload.

The Search API rate limiter uses a leaky bucket algorithm. Requests arriving faster than the leak rate accumulate in a burst bucket matching maximum burst capacity. Once that bucket fills, incoming requests are rejected with HTTP 429 status codes until earlier units leak out.

### Token Ceilings and Concurrency Budgets

For complex agentic workflows, the Agent API enforces two concurrent checks that every request must satisfy:

1. **Request Limits (QPS):** A strict limit on queries per second shared across all models and API keys within the organization.

2. **Token Limits (TPM and TPH):** Distinct tokens-per-minute and tokens-per-hour ceilings enforced on a per-model basis.

Because token budgets operate on rolling windows, a developer can receive an HTTP 429 error even when their request velocity is well below their allowed QPS limit. If earlier calls in the window ingested large document chunks, the model token allowance may be temporarily exhausted.

### Handling HTTP 429 Errors and Implementing Exponential Backoff

When an API call exceeds platform boundaries, Perplexity returns an HTTP 429 response carrying a `Retry-After` header. Inspecting the JSON response body reveals the specific failure mode:

* `request_rate_limit_exceeded`: The application exceeded its queries-per-second ceiling. The client should pause for the duration specified in `Retry-After` (typically 1 second) and retry.

* `token_rate_limit_exceeded`: The cumulative token budget (TPM or TPH) has been exhausted. The client must apply exponential backoff before retrying.

* `overloaded`: An upstream model cluster is experiencing transient capacity constraints. This is not an account rate limit; requests can be retried with jitter or routed to a fallback model.

A common implementation mistake involves high-frequency polling. In the Agent API, polling `GET /v1/agent/{id}` to check task status counts against the organization request limit. At Tier 0 (1 QPS), polling once per second leaves zero request capacity for new operations. Automated pipelines should implement exponential backoff polling intervals or use streaming connections to prevent self-inflicted throttling.

## How Can Teams Manage Large Document Repositories Beyond Upload Limits?

Engineering teams, research groups, and operational departments frequently hit hard operational boundaries when using Perplexity for extensive document analysis. Strict session attachment caps, individual Project file ceilings, and aggregate Project storage boundaries cannot accommodate large-scale research collections containing technical specifications, regulatory filings, discovery records, or massive codebases.

When teams need AI assistants to analyze repositories that exceed these boundaries, relying on native file attachments creates severe operational friction. Several architectural strategies address this limitation:

### Evaluating Document Ingestion Architectures

Organizations typically evaluate three distinct paths for handling large document collections:

1. **Manual File Splitting:** Dividing large PDFs and spreadsheets into smaller pieces to fit within upload ceilings. This method is labor-intensive, damages cross-document context, and rapidly consumes daily session upload allowances.

2. **Custom Cloud Object Storage with RAG Pipelines:** Storing raw documents in cloud object storage (such as Amazon S3 or Google Cloud Storage) and constructing a custom retrieval-augmented generation (RAG) pipeline using external vector databases like Pinecone or Weaviate. While technically flexible, building and maintaining custom chunking heuristics, embedding sync jobs, and access control layers requires dedicated engineering maintenance.

3. **Intelligent Workspaces with Native Model Context Protocol (MCP) Integration:** Offloading the entire document corpus into an intelligent cloud workspace designed for agentic retrieval, such as Fastio.

### The Intelligent Workspace Architecture on Fastio

In an intelligent workspace, document collections are stored in shared cloud workspaces where capacity easily accommodates large archives. Rather than struggling with strict session attachment caps, Fastio supports single-file uploads across plans, scaling from Starter up through Enterprise as outlined in the plan table below.

When Intelligence Mode is enabled on a workspace, incoming documents are automatically indexed for hybrid search, combining full-text indexing, semantic embeddings, and metadata filtering. AI assistants do not need documents attached directly to prompt transcripts. Instead, assistants connect to Fastio through the remote Model Context Protocol (MCP) server over Streamable HTTP at `https://mcp.fast.io/mcp/tools` (for Claude and general personal agents) or `https://mcp.fast.io/mcp/code` (for coding agents). Complete setup details are documented in the [Fastio MCP documentation](https://mcp.fast.io/docs) and [skill reference](https://mcp.fast.io/skill.md).

Through MCP tool calls, the AI assistant searches the indexed workspace dynamically, retrieving only the relevant text passages, tables, and citations required to answer the user prompt. The assistant operates with complete access to a multi-gigabyte corpus without consuming session attachment quotas, bloating prompt context windows, or risking 7-day retention deletions. Fastio leaves every vendor's own upload limit exactly where it is; what it adds is a searchable place for the files that do not fit.

### Cloud Sync Versus Static Snapshot Imports

Unlike static cloud connector imports that fail to reflect upstream file edits, Fastio provides Cloud Sync for Google Drive, Dropbox, Box, and OneDrive. Cloud Sync connects external folders to a workspace, one-way or two-way, on a schedule or on demand. When team members update source documents in Google Drive, Dropbox, Box, or OneDrive, scheduled synchronization automatically refreshes the workspace repository without requiring manual file re-uploads. SharePoint document libraries connect through the OneDrive connector.

### Structured Document Processing with Metadata Views

For complex operational documents like commercial agreements, vendor invoices, financial disclosures, and real estate packets, standard semantic search is often insufficient. Fastio provides [Metadata Views](/product/document-data-extraction/) to transform unstructured files into queryable spreadsheets.

Users describe the target fields in natural language, and the system extracts structured schemas across Text, Decimal, Integer, Boolean, Date & Time, and JSON types without requiring custom OCR rules or rigid templates. AI agents can query these structured metadata tables directly through MCP actions, retrieving exact contractual deadlines or invoice totals across thousands of documents in milliseconds.

### Workspace Permissions, Detailed Activity Logs, and Predictable Pricing

Deploying shared document workspaces across human teams and autonomous agents requires strict organizational governance. Fastio establishes role-based permissions at the organization, workspace, and share levels, ensuring agents and collaborators access only authorized [intelligent cloud workspaces](/product/workspaces/). A detailed activity log records changes, tracking uploads, modifications, moves, and deletions across all team members and connected agents. A share can also disable file downloads when distributing materials externally.

Teams can review options for [storage for agents](/storage-for-agents/) and evaluate plan structures. Monthly plans start with a 14-day free trial, which requires a credit card. Review all plan tiers on the [Fastio pricing page](/pricing/):

| Fastio Plan | Monthly Price | Annual Billing | Storage Allowance | Maximum File Upload | Included Seats | Monthly AI Credits |
| :--- | :--- | :--- | :--- | :--- | :--- | :--- |
| **Starter** | $9.99 | $99 | 250 GB | 25 GB | 3 seats | 100,000 |
| **Business** | $49.99 | $499 | 5 TB | 50 GB | 10 seats | 600,000 |
| **Enterprise** | $199.99 | $1,999 | 25 TB | 100 GB | 30 seats included (up to 200) | 3,000,000 |

Credits meter AI work, and overage is billed through standard rates. Additional storage beyond the plan allowance and additional bandwidth scale as needed.

## Frequently asked questions

### How many searches can you do on Perplexity for free?

Free accounts on Perplexity receive unmetered standard searches using default base models. For advanced multi-step reasoning, free users receive 5 Pro searches every 4 hours (amounting to roughly 30 searches across a 24-hour cycle) and 1 Research Mode query per month.

### What is the file upload limit on Perplexity Pro?

Perplexity Pro subscribers receive rolling allowances for file uploads that accommodate 50+ file uploads per day (often evaluated across a rolling weekly allowance of 500 or more files). Individual session attachments are capped at 40 MB per file (or 25 MB on certain mobile interfaces), while files stored persistently in Projects are capped at 100 MB per file.

### When do Perplexity usage limits reset?

Perplexity search and upload limits operate on rolling windows rather than a fixed calendar midnight. Standard free Pro searches reset on a rolling 4-hour cycle. Pro session file uploads reset on a rolling 7-day window, where each uploaded file stops counting against your allowance approximately 168 hours after it was attached.

### What happens when you exceed your Perplexity file upload allowance?

When an account reaches its upload limit, Perplexity displays an upgrade or quota notification when attempting to attach new files. File upload capabilities are not permanently removed. As older attachments age past the rolling window, upload capacity restores automatically.

### What is the maximum file size for Perplexity Projects?

Perplexity caps individual project file attachments across standard search sessions at 100 MB per file, with cumulative project storage capped at 5 GB and 10,000 files total. These limits apply identically across all subscription tiers, including Enterprise accounts.

### Does Perplexity sync live file updates from Google Drive or Dropbox?

No. When files are imported into Perplexity Projects through File Connectors, Perplexity creates a one-time static snapshot. Changes made to the original document in Google Drive or Dropbox do not synchronize automatically. Users must manually re-import updated files to reflect changes.

## Sources

- [Perplexity Help Center: File Uploads](https://www.perplexity.ai/help-center/en/articles/10354807-file-uploads): Perplexity caps individual file attachments at 40 MB across all supported document and media formats in standard search sessions.
- [Perplexity API Documentation: Rate Limits & Usage Tiers](https://docs.perplexity.ai/docs/admin/rate-limits-usage-tiers): Perplexity API usage tiers advance automatically based on cumulative lifetime credit purchases without demotion.

## About Fast.io

Fast.io provides shared workspaces where people and AI agents work on the same files, with built-in semantic search and citation-backed chat over what they hold. Agents reach it through a remote MCP server, a REST API at https://api.fast.io/current/, and a command line client published on npm as @vividengine/fastio-cli. MCP setup is at https://mcp.fast.io/docs: Claude and most MCP clients connect to https://mcp.fast.io/mcp/tools, ChatGPT to https://mcp.fast.io/mcp/operations, and coding agents to https://mcp.fast.io/mcp/code.