top of page
Search

Google Gemini: File upload limits and supported file types + updated capabilities

ree

Gemini’s file-handling system evolves for larger and more complex workflows.

Google Gemini now offers expanded file upload capabilities across its ecosystem, covering Gemini Apps, Vertex AI Pro and Flash, and the Gemini Files API. Each surface enforces different file size caps, formats, retention policies, and quotas, making it critical to understand these distinctions when working with research, data-heavy tasks, or multimodal content. Recent updates have significantly raised capacity limits and introduced new optimization tools for developers and enterprise users.



Upload limits across Gemini surfaces.

Gemini’s file upload handling differs depending on where the files are processed. The table below summarizes the most updated specifications across Gemini Apps, Vertex AI, and the Files API.

Surface / Model

Per-File Limit

Files per Prompt

Pages / Frames

Retention

Supported Types

Gemini Apps (Advanced)

75 MB (chat & mobile) / 100 MB (Drive-linked Docs, Slides)

10

N/A

30 days

DOC, DOCX, PDF, RTF, PPT, XLS, CSV, JPG, PNG, MP4, MP3

Gemini 2.5 Pro (Vertex AI)

50 MB

Up to 3,000

Up to 1,000 PDF pages

Prompt-only

PDF, TXT, images, audio, video

Gemini 2.5 Flash / Flash-Lite

50 MB (Flash) / 7 MB (Lite console)

Up to 3,000

1,000 pages

Prompt-only

PDF, TXT, images

Gemini Files API

5 GB per file (50 GB project limit)

No strict per-prompt limit

N/A

48 hours

Any document, image, audio, or video file

Agentspace Workspaces

100 MB

N/A

N/A

Workspace lifetime

Docs, sheets, presentations, images

These new limits represent a significant improvement over earlier restrictions, especially in the Files API, where the per-file cap increased from 2 GB to 5 GB and project storage rose to 50 GB.



How Gemini handles token consumption.

When uploading files, Gemini automatically converts content into tokens that count toward the context window. Understanding these conversion rules is key to avoiding overflow errors and optimizing performance.

Content Type

Approx. Tokens Used

Notes

Text

~4 tokens per 1 KB

Based on UTF-8 encoding

Images

~512 tokens per image

Up to 32 MP, larger images use more

Audio

~1 token per second

MP3, WAV, FLAC supported

Video frames

~120 tokens per frame

1 fps keyframes recommended

Gemini’s Vertex AI now supports context caching, allowing developers to store frequently used files and re-reference them at a 75% token discount for follow-up queries.



Free vs paid upload capabilities.

Gemini’s upload quotas and file size allowances vary significantly depending on whether you use the free Flash-Lite plan, Gemini Advanced, or Vertex AI.

Plan

Context Window

Per-File Limit

Quota

Free / Flash-Lite

128,000 tokens

7 MB per file

200 MB per rolling 7 days

Gemini Advanced (Paid)

1,000,000 tokens

75 MB (chat) / 100 MB (Docs)

1 GB daily, extra billed

Vertex AI Pro / Flash

1,000,000 tokens

50 MB per file

20 GB project quota

Gemini Files API

N/A (storage only)

5 GB per file

50 GB project storage

Paid tiers unlock larger file support, higher retention, and priority routing to Gemini Pro for multimodal tasks.



Privacy, retention, and compliance considerations.

Gemini enforces different data retention policies depending on the upload method:

  • Gemini Apps — files inherit Google Drive sensitivity labels and remain available in chat history for 30 days.

  • Vertex AI & Flash — files are stored only in-memory and are deleted immediately after completion of the prompt.

  • Gemini Files API — files are stored for 48 hours by default; automatic purge resets on access.

  • Regional data residency — enterprise users can select US, EU, or APAC storage for compliance requirements.

These features make Gemini suitable for both individual and enterprise-grade document processing, while maintaining security and governance control.



Practical tips for optimizing Gemini uploads.

To reduce token usage and avoid hitting limits, follow these strategies:

  1. Split large PDFs into ≤50 MB and ≤1,000 pages before upload.

  2. Resize images to ≤24 MP and upload no more than 10 at a time.

  3. For videos, extract 1 fps keyframes to minimize token consumption.

  4. Use the Files API to cache static data and reuse file handles across multiple calls.

  5. Track quotas using the Gemini dashboard to avoid unexpected prompt rejections.


Gemini’s roadmap for file handling.

Google is testing enterprise-only preview features that could reshape file handling:

  • Direct 2 GB uploads into prompts via the --long-upload-beta flag

  • Streaming context paging to handle inputs exceeding 1M tokens

  • Integration of Gemini’s Files API with Google Drive and Workspace for seamless document orchestration

These capabilities are currently in controlled rollout, with wider availability expected later this year.



____________

FOLLOW US FOR MORE.


DATA STUDIOS


bottom of page