Gemini Token Counter
Estimate your text in Gemini tokens and price it across all 5 Google models — every one of which accepts 1,048,576 tokens, about 1,565 pages of prose. Google does not publish Gemini’s tokenizer, so the count below is an estimate; the exact figure is a free API call away.
- Countingest.
- Context1M
- Input$2.00
- Output$12.00
- Cached$0.20010% of input
Estimate Gemini tokens
Opens on Gemini 3.1 Pro. The count is the same on every Gemini model — what changes when you switch is the price, and between the cheapest and the dearest row that is roughly 7× on input.
Drop files here or click to browse
Supports PDF, TXT, MD, JSON, CSV, XML — max 10MB
Count as an API request
Providers also bill the role and boundary tokens that wrap your text. Anthropic's own example — a short system prompt plus one message — costs 14 input tokens for 7 tokens of visible text.
How accurate is this Gemini token count?
It is an estimate. Google documents that Gemini averages about four characters per token for English but does not release the vocabulary, so no browser can reproduce its segmentation exactly. The number here is calibrated on English prose and lands close for ordinary writing; it drifts on code, JSON and scripts such as Devanagari or Chinese, where a single character can cost more than one token.
For an exact count Google exposes countTokens, which is free, does not bill against your quota and accepts the same request body as a generation call. Use it when the precision matters. It also handles something no text counter can: Gemini is natively multimodal, so images, audio, video and PDFs are converted into tokens too — an image is hundreds of tokens, a minute of video is thousands — and they draw from the same window and the same bill as your prompt.
Gemini models and prices
5 Google models, all sharing the same 1,048,576-token window and the same cache discount: cached reads cost 10% of input on every row. What separates them is price, from Gemini 3.5 Flash-Lite at $0.300 per 1M input to Gemini 3.1 Pro at $2.00.
| Model | Inputper 1M tokens | Cachedper 1M tokens | Outputper 1M tokens | Context | Counting | Your cost |
|---|---|---|---|---|---|---|
| Gemini 3.1 Pro | $2.00 | $0.200 | $12.00 | 1M | est. | — |
| Gemini 3.7 Flash | $0.750 | $0.075 | $3.75 | 1M | est. | — |
| Gemini 3.5 Flash | $1.50 | $0.150 | $9.00 | 1M | est. | — |
| Gemini 3.5 Flash-Lite | $0.300 | $0.030 | $2.50 | 1M | est. | — |
| Gemini 2.5 Pro | $1.25 | $0.125 | $10.00 | 1M | est. | — |
Rates in USD per 1M tokens, read from Google’s own pricing page on August 22, 2026. Providers change prices without notice.
Gemini’s 1M-token context window
Every Gemini model in this table accepts 1,048,576 tokens — not a round million but exactly 2²⁰, which is why the figure reads oddly. That is roughly 1,565 pages of prose at about 670 tokens a page, or a few hours of video once you count media tokens. Uniformity is unusual: on most providers the cheap tiers get a smaller window, whereas here Gemini 3.5 Flash-Lite holds exactly as much as Gemini 3.1 Pro.
The catch is at the top of the range rather than the bottom. Google bills prompts above 200,000 input tokens at roughly double the rates shown, so a request that uses the window fully costs more per token than one that half-fills it. Sending 1,048,576 tokens to Gemini 3.1 Pro is $2.10 at the standard rate before the tier adjustment, and before the model writes anything — worth checking against retrieval, which sends a fraction of the tokens for a fraction of the price.
What Gemini tokens cost
Three request shapes priced on Gemini 3.1 Pro and Gemini 3.5 Flash-Lite — the two ends of Google’s range. Note how the gap widens as the model writes more.
| Request | Gemini 3.1 Pro | Gemini 3.5 Flash-Lite |
|---|---|---|
| One chat turn1,000 in / 500 out | $0.0080 | $0.0015 |
| A long PDF100,000 in / 2,000 out | $0.224 | $0.035 |
| A full-window job1,000,000 in / 50,000 out | $2.60 | $0.425 |
Choose a Flash tier on total cost, not on the input rate. Gemini 3.5 Flash-Lite looks cheapest at $0.300 in — less than half Gemini 3.7 Flash’s $0.750 — but its output costs $2.50, more than eight times its own input rate, so anything that generates at length can end up dearer than the model that looked more expensive. Gemini 3.7 Flash is the better-balanced row: exactly half Gemini 3.5 Flash’s input rate and well under half its output.
Gemini token counter FAQ
How many tokens is 1,000 words in Gemini?
About 1,330 tokens of English prose, and 1,000 tokens comes back to roughly 752 words. Google’s own rule of thumb is about four characters per token, which lands in the same place for English. Both are averages: technical text, code and non-Latin scripts cost more per word, sometimes substantially.
What is Gemini’s context window?
1,048,576 tokens on all 5 models here — exactly 2²⁰, about 1,565 pages of prose. Unlike most providers Google does not shrink the window on its cheaper tiers, so Gemini 3.5 Flash-Lite takes the same size prompt as Gemini 3.1 Pro. Prompt, media and the generated answer all share the budget.
Is this Gemini token counter accurate?
It is an estimate, because Google does not publish Gemini’s vocabulary. Expect it to be close on English prose and looser on code, JSON and non-Latin scripts. When you need certainty, the free countTokens method returns the exact count for a request and does not bill against your quota — and it is the only way to count images, audio and video, which no text-based counter can see.
Do images and PDFs count as Gemini tokens?
Yes. Gemini is natively multimodal, so every attachment is tokenized: an image is worth hundreds of tokens, a minute of video thousands, and a PDF is charged for both its text and its page images. They come out of the same 1,048,576-token window and the same bill as your prompt. The counter above measures text only — for documents, use the file counter, which extracts and counts the text of a PDF or Word file in your browser.
How much does Gemini 3.1 Pro cost?
$2.00 per 1M input tokens, $12.00 per 1M output and $0.200 for cached reads — 10% of the input rate, the same ratio Google applies to every model in this table. Watch the size tier: above 200,000 input tokens Google bills the request at roughly double these figures.
Should I use Gemini Flash or Flash-Lite?
It depends on how much the model writes. Gemini 3.5 Flash-Lite is the cheapest Google model on both rates — $0.300 in, $2.50 out — but that output figure is more than eight times its own input rate, an unusually steep ratio, so a generation-heavy workload narrows the gap fast. Gemini 3.7 Flash at $0.750 / $3.75 is the more balanced choice, and it undercuts Gemini 3.5 Flash on both rates. Put your real token volumes into the calculator above and the table re-prices around them.
Gemini against the field
A 1,048,576-token window is generous but not unique — several providers now match it, at very different prices. Compare 56 models from 11 providers, or open another brand’s counter below.
New to tokens? Read how tokenization works