Gemini's context window in the app is 32k tokens without an AI plan, 128k tokens on Google AI Plus and 1 million tokens on Google AI Pro and AI Ultra, as of October 2026. Google says a 1 million token window lets Gemini understand up to 1,500 pages of text or 30,000 lines of code at once.
Google's own analogy is that the context window is Gemini's reading capacity: how much it can take in together at one time. If you want the general idea behind the term, my explainer on context windows covers it. This page is about Gemini's numbers, what they mean in practice, and why the figure a developer quotes might not match yours.
Gemini app context window by plan
| Plan or account | Context window, as of October 2026 |
|---|---|
| Gemini without an AI plan | 32k tokens |
| Google AI Plus | 128k tokens |
| Google AI Pro | 1 million tokens |
| Google AI Ultra | 1 million tokens |
| Work account, lowest Workspace tier | 32 thousand |
| Work account, higher Workspace tiers | 1 million |
Two things jump out. Ultra doesn't give you a bigger window than Pro; the jump to 1 million happens at Pro. And on a work account, the lowest Workspace tier has the same window as the free personal app, so a team on a basic licence isn't getting the long-document capacity the marketing implies. If you're deciding on a plan for this reason, my guide to Google AI Pro covers what else comes with it.
What a million tokens holds
Google gives two sets of examples, one in the Gemini help centre and one in its developer docs. The help centre says 1 million tokens is up to 1,500 pages of text or 30,000 lines of code. The developer docs describe it like this:
- 50,000 lines of code, at the standard 80 characters per line
- All the text messages you've sent in the last 5 years
- 8 average length English novels
- Transcripts of over 200 average length podcast episodes
Spot the mismatch: 30,000 lines of code on one Google page, 50,000 on the other. Neither is wrong as such; they're rough illustrations, not guarantees. I'd treat both as a feel for scale and never plan a job right up to the edge.
For perspective, the same developer page says earlier models could only handle 8,000 tokens at a time, then 32,000 or 128,000. So the free plan's 32k isn't tiny. It's what counted as large not long ago.
Why the API window is a different number
If a developer tells you Gemini handles a million tokens, they're usually talking about the Gemini API, not the app. The API rules differ by model. In Gemini 4 Argon, the 1 million token window is shared across input, thinking and output tokens. If your input plus the maximum output you ask for goes over 1,048,576 tokens, the API automatically caps the output to whatever room is left. Earlier Gemini 3 and 2.5 models used separate budgets: 1 million tokens in, 64K out.
None of that changes what you get in the app. In the app, your window depends on your plan, as in the table above. Keep the two separate in your head and half the confusing conversations about Gemini's limits go away.
Why long chats get slower and weaker
Google explains what happens as a chat grows. In an ongoing conversation, Gemini has to re-read the previous history to keep its context, so the longer the chat, the more thinking power each reply takes, and the more of your usage it eats. Large file analysis works the same way; Google's example is asking Gemini to analyse a 100-page PDF financial report.
Go past the window and the quality drops in a specific way. Google says answers may not take into account all the content you gave it, or may miss connections and details across it. That's most noticeable when the details you care about are scattered through a large file. Limits refresh every 5 hours until you hit a weekly limit, and my guide to Gemini usage limits goes into how that works.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
How I'd work within Gemini's context window
- Start a new chat when you move to a new topic. That's Google's own tip, because Gemini won't have to re-read a long history first.
- Open each project chat with a short brief: who it's for, what good looks like, the key facts. Then you can start fresh without losing the thread.
- On the free plan, upload the sections of a long document you need, not the whole thing.
- If you regularly analyse long reports or contracts, that's where a 1 million token window starts to earn its place.
- When details are spread through a big file, check the answer against the source yourself, because that's exactly where Google says misses happen.
My verdict: I'd start a fresh chat for each project and skip the one endless thread that's been running since spring. A bigger window helps, but a cleaner chat helps more, on every plan. Context sizes change often, so check Google's own help page for the latest.
If your team keeps running into the limits of long AI chats, that's the kind of AI training I run. I run AI training and workshops for teams, and help businesses with AI enablement and workflow automation, so the busywork runs itself and people get time back. Most of that work is with SaaS marketing teams and small businesses. Here's how we can work together.
Frequently asked questions
What is Gemini's context window?
As of October 2026, the Gemini app has a 32k token context window without an AI plan, 128k on Google AI Plus and 1 million tokens on Google AI Pro and AI Ultra.
How much is 1 million tokens in Gemini?
Google says it's up to 1,500 pages of text or 30,000 lines of code. Its developer docs also describe it as 8 average length English novels or transcripts of over 200 average length podcast episodes.
Does Google AI Ultra have a bigger context window than Pro?
No. Google lists both AI Pro and AI Ultra at 1 million tokens in the Gemini app.
Is the Gemini API context window the same as the app?
Not necessarily. In the API, Gemini 4 Argon shares its 1 million token window across input, thinking and output, while in the app your window depends on your plan.
Why does Gemini forget things in long chats?
Google says Gemini re-reads the conversation history in ongoing chats, and if content goes past the context window, answers may miss details or connections. Starting a new chat for a new topic helps.
