Gemini file upload limits in 2026: 100 MB per file, 2 GB video, 10 files per prompt

Updated: November 2, 2026

Gemini Apps accept up to 10 files in one prompt, with each video up to 2 GB and every other supported file up to 100 MB. Google's Gemini Apps Help page also caps total video length at 5 minutes and total audio length at 10 minutes for users without a paid plan, rising to 1 hour of video and 3 hours of audio on Google AI Pro and Google AI Ultra.

The size cap is only the first of three limits a Gemini user meets. The second is the context window, which Google lists as 32,000 tokens without a plan, 128,000 tokens on Google AI Plus and 1 million tokens on Google AI Pro and Ultra. The third is a rolling usage limit on how much content can be uploaded and analyzed over time. Below are the full limit table, the plan differences, the developer limits of the Gemini API, a page-count calculation for long PDFs and the steps that avoid failed uploads.

........

Gemini Apps upload item	Current limit
Files per prompt	Up to 10 (subject to availability)
Video file size	Up to 2 GB each
Any other supported file	Up to 100 MB each
Total video length, no paid plan	Up to 5 minutes
Total video length, AI Pro or Ultra	Up to 1 hour
Total audio length, no paid plan	Up to 10 minutes
Total audio length, AI Pro or Ultra	Up to 3 hours
Code folder or GitHub repository	One per chat, up to 5,000 files and 100 MB
ZIP archive	Up to 10 files and 100 MB, no audio or video inside

........

··········

GEMINI APPS FILE UPLOAD LIMITS FOR DOCUMENTS, MEDIA AND CODE.

Google describes Gemini Apps as supporting "most file types" and does not publish a closed list of formats for the consumer app. Its help page sets specific rules for video, audio, code folders, GitHub repositories and ZIP archives, and it mentions spreadsheets, which the Gemini web app can turn into charts; every other supported file falls under the general 100 MB rule. When a folder is uploaded, every file inside it counts toward the 10-file limit of the prompt.

Video is the only type with a larger size ceiling. A single video file can reach 2 GB, while every other supported file stops at 100 MB. Duration is a separate cap that applies to the total of all videos or all audio files in the upload, so two 4-minute clips already exceed the 5-minute video allowance of an account without Google AI Pro or Ultra.

Code has its own rule. One code folder or one GitHub repository can be added to a chat, with up to 5,000 files and a maximum size of 100 MB. ZIP archives follow a stricter pattern: up to 10 files and 100 MB in total, and Google states that a ZIP cannot include video or audio files.

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣

GEMINI CONTEXT WINDOW BY PLAN AND WHY A VALID FILE CAN STILL FAIL.

An accepted upload is not the same as a fully read upload. Gemini can only reason over the amount of text, image and media data that fits in the context window of the plan and model in use, and Google warns that very large files can lead Gemini to miss details or connections spread across the material.

........

Google plan	Gemini Apps context window	Usage limit compared with no plan
No AI plan	32,000 tokens	Standard limits
Google AI Plus	128,000 tokens	2x higher
Google AI Pro	1 million tokens	4x higher
Google AI Ultra	1 million tokens	5x or 20x higher than AI Pro, by subscription

........

Google states that a 1 million token context window lets Gemini understand up to about 1,500 pages of text or 30,000 lines of code. Deep Think mode works with a smaller window of 192,000 tokens, according to the same help page. Usage limits refresh every 5 hours until the user reaches a weekly limit, and Google says these limits may change without notice; the Usage Limits page in Gemini settings shows the figures that apply to each account.

A free account with a 100 MB PDF of several hundred pages is the most common case of a file that uploads without error and still produces incomplete answers, because the document holds far more text than 32,000 tokens.

··········

HOW MANY PDF PAGES FIT IN GEMINI: A TOKEN CALCULATION.

Google does not publish a per-page token figure for the consumer app, but the Gemini API documentation does: each PDF page processed through document vision counts as 258 tokens. Dividing each context window by that figure gives a rough upper bound on how many PDF pages could fit before the prompt, the answer and any other files take their share. The calculation was run in Python.

........

Context window	Size in tokens	PDF pages at 258 tokens per page (API rate)
No AI plan (Gemini Apps)	32,000	About 124
Google AI Plus (Gemini Apps)	128,000	About 496
Deep Think mode	192,000	About 744
Google AI Pro and Ultra (Gemini Apps)	1,000,000	About 3,875

........

These are Data Studios estimates based on an API rule and should be read as an order of magnitude for the app. Google's own consumer figure for the 1 million token window is about 1,500 pages of text, a more conservative number that reflects dense text pages, the reply and the conversation history. The two figures agree on the practical point: on a free account, a long report or a book-length PDF should be split before upload.

··········

[[ADPLUS_300x250_1]]

GEMINI API FILE LIMITS FOR DEVELOPERS: INLINE DATA, FILES API AND PDFS.

Developers who call Gemini through the API work under different, documented limits. The Gemini API file input methods page, last updated on October 7, 2026, compares four ways of sending files.

........

Gemini API input method	Maximum size	How long the file is kept
Inline data	100 MB per request (50 MB for PDFs)	Not stored, sent with every request
Files API upload	2 GB per file, 20 GB per project	48 hours
Files API with Cloud Storage URI	2 GB per file, no overall storage limit	Fetched per request, registration valid up to 30 days
External URL	15 MB per file	Fetched per request

........

The document understanding page, last updated on October 8, 2026, sets the PDF limit at 50 MB or 1,000 pages, a cap that applies to both inline data and Files API uploads. Pages above 3072 x 3072 pixels are scaled down and smaller pages are scaled up to 768 x 768. Google adds that document vision only meaningfully understands PDFs: TXT, Markdown and HTML files are read as plain text, so charts, diagrams and formatting in those formats are lost.

Teams that build document pipelines in Google AI Studio can compare these rules with our guide to Google AI Studio PDF and file upload limits.

··········

ROLLING LIMITS, DRIVE FILES AND WORK ACCOUNTS IN GEMINI.

Google applies rolling limits to how much a user can upload and analyze in a period of time. Once the limit is reached, uploads stop working until capacity resets; Google does not publish a fixed number of files per day, and upgrading to a higher Google AI plan raises the overall usage allowance.

Files can also come from Google Drive or NotebookLM. Adding Drive files requires Keep Activity to be on and Google Workspace to be connected to Gemini Apps, and Google notes that files from Drive or NotebookLM take more of the separate storage allowance that Gemini Apps keeps for chats and uploaded files. On a work or school account, the Workspace administrator must enable access to apps in Gemini before Drive files can be used.

When Gemini fails to analyze a file correctly, Google's advice is to upload it again. A new upload does not reset a rolling limit, and our guide to Gemini file upload errors separates analysis failures from limit errors and context overflow.

··········

⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁤

HOW TO UPLOAD LARGE FILES TO GEMINI WITHOUT LOSING DETAIL.

The first step is to match the file to the plan. On an account without a paid plan, documents of more than roughly a hundred dense pages exceed the 32,000-token window, so splitting by chapter or section gives better answers than one large upload. On Google AI Pro or Ultra, entire reports, contracts or code folders usually fit in one conversation.

The second step is to remove material the question does not need. Appendices, scanned images of text and duplicated tables add tokens without adding answers, and scanned pages are harder to read than text-native PDFs. For PDFs with many tables, our article on Gemini PDF extraction for text, tables and structured outputs describes a schema-based approach that keeps results consistent.

The third step concerns media. Trimming a video to the relevant segment keeps it inside the 5-minute free allowance, and converting a long recording to a shorter excerpt does the same for the 10-minute audio cap. Archives should never carry media files, because ZIP uploads reject audio and video.

··········

GEMINI FILE UPLOAD LIMITS: WHAT GOOGLE PUBLISHES AND WHAT IT DOES NOT.

Google publishes the hard caps of Gemini Apps clearly: 10 files per prompt, 100 MB per file, 2 GB per video, 5 minutes of video and 10 minutes of audio without a paid plan, 1 hour and 3 hours with AI Pro or Ultra, and dedicated limits for code folders and ZIP archives. It also publishes the context window of each plan, from 32,000 tokens without a plan to 1 million on AI Pro and Ultra.

Google does not publish a complete list of supported formats for the consumer app, a per-page token rate for the app, or a numeric daily quota for uploads, which it manages through rolling and weekly usage limits that can change without notice. For developers, the Gemini API documents every figure, from the 100 MB inline limit to the 2 GB Files API cap, the 48-hour retention and the 50 MB or 1,000-page PDF ceiling. All figures in this guide were checked on Google's help and developer pages on November 2, 2026.

··········

FOLLOW US FOR MORE.

·····

DATA STUDIOS

·····

datastudios.org