Gemini Token Counter
Token estimates for Google's Gemini models, plus the image, video and audio rules.
How do you count Gemini tokens? Google says a Gemini token is about 4 characters, and 100 tokens is about 60 to 80 English words. This counter divides by 3.7 characters per token. Against Gemini's own countTokens method, that read 4% high on a technical article and 14% high on a novel, and low on code and JSON. For an exact count, call countTokens with your API key.
$2 input · $10 output per 1M tokenspromo to 30 Nov
- Input tokens (est.)
- 0 · $0.00000
- Output cost
- $0.0050
Same text on other models
Runs in your browser. OpenAI models show exact counts from the official tiktoken tokenizer (o200k_base). Other providers are estimates from characters-per-token ratios, so treat them as a guide; for exact counts use Anthropic's count_tokens endpoint or Gemini's countTokens method. Prices shown are standard rates; some models charge more for very long prompts (for example OpenAI GPT-5.5 and later above 272K input tokens, xAI above 200K, Claude Haiku 5.5 above 100K).
How many tokens do images, video and audio use in Gemini?
Gemini tokenizes everything you send, not only text. Google's token guide gives fixed rates: an image no larger than 384 pixels on both sides counts as 258 tokens, and bigger images are cut into 768 by 768 pixel tiles at 258 tokens each. Video costs 263 tokens a second when processed as static frames, and audio 32 tokens a second.
Those rates add up fast. A one-minute audio clip is 1,920 tokens, ten minutes is 19,200, and a one-minute video is 15,780, all before the text of your prompt. Count the media first when you are close to a model's context window or a daily budget.
Why does this estimate differ from Google's rule of thumb?
Google's 4 characters per token is an average. We checked our 3.7 against Gemini's countTokens method on five samples, written up in our test of how many tokens a word is. English and Spanish prose came to 3.8 to 4.2 characters per token, a TypeScript file to 3.2 and JSON to 2.5. So the estimate reads a little high on prose and low on code: 15% low on the TypeScript file and 33% low on JSON.
A prompt that is mostly instructions with some code lands close to 3.7, which is why we use it. Gemini 3.6 Flash and 3.5 Flash Lite matched 3.8 Flash exactly where we reran the same text, and Gemini 4 Argon was within 2%, so one ratio covers them all. Paste mostly code or JSON and expect the real count to be higher than the estimate.
When the number has to be exact, call countTokens on the model you will use, or read usage_metadata in a real response. Both need an API key; the estimate on this page does not, and your text never leaves the browser.
What does a Gemini prompt cost?
Gemini 3.8 Flash charges $0.75 per 1M input tokens and $3.75 per 1M output tokens on a promotional rate that ends 31 Dec 2026, then $1.50 / $7.50. Every Gemini model's rate is in our Gemini API prices table, and the LLM cost calculator turns a token count into a monthly bill. Promo end dates and thinking-token billing are on the Gemini API pricing page. Many Gemini models can also be used for free with limits shown in AI Studio; see the Gemini API free tier for the catch.
Frequently asked
How many words is 100 Gemini tokens?
About 60 to 80 English words, according to Google's token guide. Put the other way, 1,000 words is roughly 1,250 to 1,670 tokens. When we measured how many tokens a word is with the countTokens method, Gemini 3.8 Flash came to 1.29 tokens per word on English fiction and 1.52 on a technical article. Code and JSON take more, so paste a sample to see the estimate for your own text.
How many tokens is an image in Gemini?
258 tokens for an image up to 384 pixels on both sides. Larger images are split into 768 by 768 pixel tiles, and each tile counts as 258 tokens, so a big screenshot can cost several times that.
Does this Gemini counter send my text to Google?
No. The estimate is simple arithmetic done in your browser, so nothing is uploaded and no API key is needed. Only Google's own countTokens method gives an exact figure, and that does send the text to the Gemini API.