Grok: file upload limits and supported formats explained

Grok’s latest capabilities have expanded significantly, allowing users to upload and analyze a wider variety of files directly in chat, via the API, or through Grok 4’s new document-processing features. From images and PDFs to code files and ZIP archives, Grok now integrates file-based intelligence into real workflows while maintaining strict data security. This guide explains the supported formats, upload limits, enterprise integration options, and the most important changes as of August 2025.

Grok expands file upload capabilities for images and documents.

The most recent Grok 4 update introduces broader file support, letting users upload documents, images, and structured data files for analysis, summarization, and extraction.

• Image formats supported: JPEG, PNG, GIF, WebP.

• Document formats supported: PDF, DOCX, TXT, MD, CSV.

• Other supported files: ZIP archives (scanned for security).

• Unsupported formats: TIFF, proprietary RAW formats, executables, and encrypted files.

Practical scenarios include uploading a PDF annual report for extraction, attaching screenshots for defect counting, or parsing compressed datasets through Grok’s integrated analysis tools.

Updated limits set new boundaries for uploads.

Grok’s file handling rules vary slightly between web chat and API endpoints, giving flexibility to developers while keeping performance predictable for regular users.

Upload Surface	Maximum File Size	Number of Files per Request	Supported Types
Web Chat	Up to 25 MB per file	3 images max	JPEG, PNG, GIF, WebP
Grok API / Grok 4	Up to 30 MB per file	10 files max	PDFs, DOCX, TXT, CSV, MD, images, ZIP
ZIP archive scanning	30 MB uncompressed	1 per request	Automatically scanned before processing

Why this matters: Larger file allowances in the Grok 4 API are particularly relevant for enterprise deployments, enabling teams to analyze batches of reports, documents, and images without breaking them into separate uploads.

Image analysis brings deeper vision capabilities.

Grok’s vision models handle images natively, enabling users to extract structured insights or ask contextual questions directly on uploaded images.

Example queries:

• “How many defects are visible in this manufacturing photo?”

• “Summarize the main findings from this lab slide.”

• “Count the labeled components in this schematic.”

This vision-powered Q&A works in both web chat and API-driven pipelines, but the image count per request is capped at 3 images in chat and 10 images through the API to balance performance and latency.

Expanded file handling comes with built-in security controls.

All files uploaded to Grok are automatically scanned for security and processed in memory only, ensuring no persistent retention unless explicitly requested by the user.

File Type	Security Treatment	Retention Policy
Images	Processed locally, never stored.	Deleted immediately after response.
Documents	Processed in session memory; no server-side persistence.	Deleted after task execution.
ZIP files	Scanned automatically for viruses or encrypted payloads; malicious archives are blocked.	Not stored beyond active session.

These measures are particularly important for regulated industries, where handling of confidential documents must comply with strict security and retention policies.

Cloud storage integration extends enterprise workflows.

For organizations managing high-volume document analysis, Grok 4 includes a cloud file integration feature now in private beta for enterprise tenants:

• Google Drive and OneDrive connectors allow users to import documents directly from existing cloud repositories.

• Files accessed through connectors respect the same security and size constraints applied to direct uploads.

• Availability is currently limited to enterprise allow-lists, with broader rollout expected later in 2025.

For businesses with highly distributed datasets, this integration minimizes redundant transfers while maintaining centralized storage governance.

Privacy-first processing and regional compliance.

Grok’s file analysis engine was designed with data minimization at its core:

• Uploaded files are processed in memory only and are not stored by default.

• Session data is erased immediately after the task completes.

• For enterprise accounts, region-specific processing ensures files remain in designated geographic boundaries when required for compliance (e.g., GDPR or CCPA scenarios).

This architecture allows organizations to leverage Grok’s analytics without compromising sensitive information or breaching data-residency policies.

Common scenarios where Grok file uploads accelerate work.

Use Case	Uploaded File Type	Grok Feature Used	Outcome
Financial report extraction	PDFs	Grok 4 API + structured output	Key metrics returned as JSON for analysis.
Defect detection	PNG or WebP images	Vision-powered image Q&A	Automated defect counts across uploaded sets.
Code review	TXT, MD, ZIP	API with chain calls	Analyzes codebase quality metrics instantly.
Batch summarization	DOCX, CSV, PDF	API endpoint	Returns summaries for hundreds of documents at once.
Data archiving	ZIP archives	Security scan + ingestion	Safely indexes compressed datasets for retrieval.

By combining vision, document parsing, and bulk API integration, Grok’s file support accelerates workflows across research, finance, software development, and operations.

Key takeaways.

• Grok 4 supports JPEG, PNG, GIF, WebP, PDFs, TXT, CSV, DOCX, MD, and ZIP formats.

• File size limits: 25 MB per file in chat, 30 MB for Grok 4 API requests.

• Image counts: Up to 3 images in chat, 10 images via API.

• Cloud connectors for Google Drive and OneDrive are in enterprise beta.

• Files are processed in memory and automatically scanned for security before analysis.

• Uploaded files are not retained unless the user explicitly enables storage.

With these expanded capabilities, Grok has evolved into a robust multimodal platform that combines advanced file handling with vision analysis and structured document parsing—while maintaining the privacy, security, and control that enterprise users expect.

____________

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org