top of page

Google Gemini context window 2026: 32K free, 128K Plus, 1M on Pro and Ultra

Dec 15, 2025
7 min read

Updated: Nov 11

Google Gemini context window 2026 cover: 32K free, 128K Plus, 1M on Pro and Ultra. Data Studios original cover

Updated: November 11, 2026


The Gemini app context window depends on your plan: Google's help center lists 32K tokens without an AI plan, 128K tokens on Google AI Plus and 1 million tokens on Google AI Pro and Google AI Ultra. Deep Think, the Ultra reasoning mode, runs with a smaller window of 192K tokens, and Google says a 1 million token window holds up to 1,500 pages of text or 30,000 lines of code.


Developers get a different set of numbers. On the Gemini API and Google Cloud, the current Gemini 4 Argon, Gemini 4 Carbon, Gemini 3.8 Flash and Gemini 3.5 Flash-Lite models all have a context window of 1,048,576 tokens, with output per response capped at 65,536 tokens on the Flash models and on streamed Gemini 4 replies. The tables below put every plan and model side by side, with file upload limits, the page and code capacity of each window and a worked token budget.


........


Gemini access

Context window

Monthly price (US)

Gemini app, no AI plan

32K tokens

$0

Google AI Plus

128K tokens

$4.99

Google AI Pro

1 million tokens

$19.99

Google AI Ultra

1 million tokens

from $99.99

Deep Think mode (Ultra)

192K tokens

included in Ultra

Gemini API (current models)

1,048,576 tokens

pay per token


........


··········


GEMINI APP CONTEXT WINDOW BY GOOGLE AI PLAN.


Google publishes the context window of the Gemini app for personal Google Accounts in its "Gemini Apps limits" help article. A user with no subscription gets 32K tokens, a Google AI Plus subscriber gets 128K tokens, and both Google AI Pro and Google AI Ultra subscribers get 1 million tokens. Ultra therefore buys more usage and access to more models, while its context window is the same as Pro's.


The plans also differ in the models they open up. Google's subscription page lists access to a Flash-Lite model for free users, the Flash model for Google AI Plus, Gemini 4 Argon together with the Pro and Flash models for Google AI Pro, and the highest access to Gemini 4 Argon, Gemini 4 Carbon and the Flash model for Google AI Ultra. Gemini 4 Argon reached the Gemini app for Pro and Ultra subscribers on October 13, 2026, according to Google's release notes.


Usage limits are separate from the context window. Google says the Gemini app limit refreshes every 5 hours until a weekly limit is reached, and it describes the paid tiers in relative terms: Plus gives 2x the standard limits, Pro 4x, and Ultra 5x or 20x the Pro limits depending on whether the subscription costs $99.99 or $199.99 a month. Subscribers who reach their limit can keep chatting with Flash-Lite. Our guide to Google Gemini usage limits, daily caps and API quotas covers those caps in more depth.


........


Plan

Context window

Models listed by Google

Usage vs free

No AI plan

32K tokens

Flash-Lite

standard limits

Google AI Plus

128K tokens

Flash

2x

Google AI Pro

1 million tokens

Gemini 4 Argon, Pro, Flash

4x

Google AI Ultra $99.99

1 million tokens

Gemini 4 Argon, Gemini 4 Carbon, Flash

5x the Pro limits

Google AI Ultra $199.99

1 million tokens

Gemini 4 Argon, Gemini 4 Carbon, Flash

20x the Pro limits


........


Deep Think works with a different budget. Google's file upload help page states that Deep Think has a 192 thousand token window, so a Pro-sized document that fits in a normal Ultra chat can be too large for a Deep Think request.


··········


⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣

GEMINI CONTEXT WINDOW FOR WORK AND SCHOOL ACCOUNTS.


Work and school accounts follow a separate table in Google's "Use Gemini Apps with a work or school Google Account" article. Editions that include the Gemini app as an additional service (Android Enterprise, Cloud Identity, Chrome Enterprise, Business Base, Essentials Starter, G Suite and Workspace Individual) get a context size of 32 thousand tokens, and editions with AI Ultra Access get 1 million tokens.


For the core-service editions, such as Business Standard, Business Plus, Enterprise Standard and Enterprise Plus, and for AI Expanded Access, the same table leaves the context size blank, so Google does not publish a figure for them. The article does publish prompt limits for those editions: up to 25 Pro prompts per 4 hours on Business Standard, Business Plus and the Enterprise Standard and Plus editions, up to 200 Gemini 4 Argon prompts a day with AI Expanded Access, and up to 500 a day with AI Ultra Access. Google adds that these limits may change without notice.


··········


GEMINI API CONTEXT WINDOWS AND OUTPUT CAPS BY MODEL.


The model pages of Google Cloud's Gemini Enterprise Agent Platform list the token limits of each model. All four current production models share a context window of 1,048,576 tokens, and they differ in how output is capped and in release date.


........


Model (API ID)

Context window

Output per response

Release date

Gemini 4 Argon (gemini-4-argon)

1,048,576

65,536 streaming, 32,768 non-streaming

October 13, 2026

Gemini 4 Carbon (gemini-4-carbon)

1,048,576

65,536 streaming, 32,768 non-streaming

November 10, 2026

Gemini 3.8 Flash (gemini-3.8-flash)

1,048,576

65,536

September 2, 2026

Gemini 3.5 Flash-Lite (gemini-3.5-flash-lite)

1,048,576

65,536

July 21, 2026


........


For the two Gemini 4 models, Google describes the 1,048,576-token limit as unified: input and output tokens share it. A single streamed segment produces up to 65,536 output tokens, and longer outputs are possible through long decodes on the v1beta1 endpoint or the Interactions API, up to 1,048,576 tokens minus the input. Gemini 4 Argon and Gemini 4 Carbon accept up to 3,000 images and documents per prompt, documents of up to 3,000 pages, up to 10 videos and about 8.4 hours of audio. Developers who need pricing, caching and the full model list can read our guide to Gemini API context windows, output caps and model differences.


··········


[[ADPLUS_300x250_1]]

GEMINI FILE UPLOAD LIMITS THAT SIT ON TOP OF THE CONTEXT WINDOW.


A large context window does not lift the per-file rules of the Gemini app. Google's help page on uploading files lists up to 10 files per prompt, videos of up to 2 GB each and up to 100 MB for every other supported file type. A code folder or a GitHub repository can be added to a chat with up to 5,000 files and 100 MB in total, and ZIP archives take up to 10 files and 100 MB without video or audio.


Time limits for media depend on the plan. Without a Pro or Ultra subscription, the total video length is up to 5 minutes and the total audio length up to 10 minutes; Google AI Pro and Ultra extend those totals to 1 hour of video and 3 hours of audio. For document-heavy work, our guide on whether Gemini can read PDFs and how its document limits work goes through accuracy and page handling.


........


Gemini app upload rule

Without Pro or Ultra

Google AI Pro or Ultra

Files per prompt

up to 10

up to 10

Size per video file

2 GB

2 GB

Size per other file

100 MB

100 MB

Total video length

5 minutes

1 hour

Total audio length

10 minutes

3 hours

Code folder or GitHub repo

5,000 files, 100 MB

5,000 files, 100 MB


........


··········


HOW MUCH TEXT FITS IN EACH GEMINI CONTEXT WINDOW.


Google gives one conversion: a 1 million token window holds up to 1,500 pages of text or 30,000 lines of code. Applying that same ratio to the smaller windows gives a rough idea of what each plan can keep in view at once. These are estimates derived from Google's figure, and real documents vary with layout, language and the share of tables or images.


........


Window

Plan or mode

Pages of text (estimate)

Lines of code (estimate)

32K tokens

No AI plan

about 48

about 960

128K tokens

Google AI Plus

about 192

about 3,840

192K tokens

Deep Think

about 288

about 5,760

1 million tokens

Google AI Pro and Ultra

up to 1,500

up to 30,000


........


The window also has to hold the conversation itself. On the API, a request that sends 900,000 tokens of documents to Gemini 3.8 Flash uses about 86% of the 1,048,576-token window and leaves 148,576 tokens, which is more than the 65,536-token output cap, so the full output budget stays available. If the input rises above 983,040 tokens, the remaining room drops below 65,536 tokens and the model can no longer return a maximum-length reply.


··········


⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁣⁤

HOW TO WORK WITHIN THE GEMINI CONTEXT WINDOW.


In the Gemini app, the simplest check is the plan itself: a free account keeps roughly a long report in view, Plus handles a short book, and Pro or Ultra can hold a large contract set or a mid-sized codebase. When a chat starts to lose track of early material, starting a new chat with only the documents that matter for the next question frees the window again.


Splitting large jobs keeps each request inside the budget. Long PDFs can be divided by chapter or section, with page numbers or section labels kept in the text so the answers can point back to the source. For code, adding a single folder or repository instead of loose files respects the 5,000-file and 100 MB limits and keeps the structure readable for the model.


On the API, the safest habit is to size each request with the reply in mind, so that input plus expected output stays under 1,048,576 tokens. A request planned for a full 65,536-token answer should keep its input under 983,040 tokens, and requests that only need a short answer can use almost the whole window for source material.


··········


GOOGLE GEMINI CONTEXT WINDOW IN 2026: CONFIRMED LIMITS AND GAPS.


Google's own pages confirm a clear ladder for personal accounts: 32K tokens free, 128K on Google AI Plus, 1 million on Google AI Pro and Ultra, and 192K for Deep Think. The API side is uniform, with 1,048,576 tokens on Gemini 4 Argon, Gemini 4 Carbon, Gemini 3.8 Flash and Gemini 3.5 Flash-Lite, and an output cap of 65,536 tokens per response or streamed segment.


Two areas remain undocumented. Google does not state a context size for the core Workspace editions or for AI Expanded Access, and it describes Gemini app usage limits for personal accounts only as multiples of a standard limit, with no published prompt count. Readers on those plans should treat the 32K and 1 million figures as the published floor and ceiling and test their own documents before relying on a single very long chat.


··········


FOLLOW US FOR MORE.


·····


DATA STUDIOS


·····


datastudios.org


bottom of page