★ Gemini · token counter
Gemini Token Counter
Google runs a free countTokens endpoint for Gemini. When this site's server has a Gemini key configured, your text is sent to it and you get Google's exact count per model. Without it, you get a labelled estimate from Google's own rule of thumb.
- Runs in your browser
- No signup
- 306
- Characters
- 54
- Words
- —
- Tokens / word
Tokens per word uses the o200k_base count. Characters count emoji and accented letters as one each.
Tokens
77
Estimated with Google's rule of thumb: about 4 characters per token.
Context window used
<0.1%
1.05M tokens window on Gemini 4 Argon.
Input cost per call
$0.00015
Per 1,000 calls
$0.1540
Every model, same text
| Model | Tokens | Accuracy | Input / call |
|---|---|---|---|
| Claude Fable 5.1Anthropic | … | ≈ Estimate | — |
| Claude Opus 5.5Anthropic | … | ≈ Estimate | — |
| Claude Sonnet 5.5Anthropic | … | ≈ Estimate | — |
| Claude Haiku 5.5Anthropic | … | ≈ Estimate | — |
| GPT-6 AstraOpenAI | … | ✓ Exact | — |
| GPT-6.1 SolOpenAI | … | ✓ Exact | — |
| GPT-6 LunaOpenAI | … | ✓ Exact | — |
| Gemini 4 ArgonGoogle | 77 | ≈ Estimate | $0.00015 |
| Gemini 3.8 FlashGoogle | 77 | ≈ Estimate | $0.00006 |
| Gemini 3.1 Flash-LiteGoogle | 77 | ≈ Estimate | $0.00002 |
| Mistral Large 4Mistral | … | ≈ Estimate | — |
| Mistral Small 4Mistral | … | ≈ Estimate | — |
| DeepSeek V4.1 FlashDeepSeek | 92 | ≈ Estimate | $0.00003 |
| DeepSeek V4 ProDeepSeek | 92 | ≈ Estimate | $0.00012 |
| Llama (open weights)Meta | … | ≈ Estimate | — |
OpenAI rows use o200k_base, exact for that tokenizer. tiktoken hasn't mapped GPT-6 model names yet, so OpenAI hasn't confirmed GPT-6 uses it. Input cost is list price for your text alone, before system prompts, caching or output.
See the tokens
o200k_base
Loading the tokenizer…
77 tokens on Gemini 4 Argon.
Prompts eating your context window? I build RAG and prompt pipelines that use a fraction of the tokens.
How Gemini tokenizes
Gemini uses its own tokenizer, which Google doesn't ship as a small browser library. The Gemini API exposes a countTokens method that returns totalTokens for a given model and input.
Google's docs give the rule of thumb: about 4 characters per token, and 100 tokens is roughly 60 to 80 English words. Images, audio and video have their own token rates.
Exact or estimate?
Exact when the server key is set: the number comes from Google's countTokens endpoint for that model. Your text is sent to Google for counting and is not stored by this site.
Estimate otherwise: characters divided by 4, Google's published rule of thumb. The page labels which one you are seeing.
Current Gemini models
Context windows and input prices from the pricing data used across these tools.
| Model | Context window | Input / 1M |
|---|---|---|
| Gemini 4 Argon | 1.05M tokens | $2 |
| Gemini 3.8 Flash | 1.05M tokens | $0.75 |
| Gemini 3.1 Flash-Lite | 1.05M tokens | $0.25 |
Quirks worth knowing
- Google documents images up to 384 px on both sides as 258 tokens, with larger images tiled into 768 x 768 tiles of 258 tokens each.
- System instructions and tool definitions count toward input tokens too.
- Non-English text and code usually differ from the 4-characters rule, which is why the exact endpoint matters.
Sources: Gemini API: Understand and count tokens · Gemini API: Gemini 4 Argon · Gemini API: Gemini 3.1 Flash-Lite
Questions people actually ask
Is the Gemini count exact?
Is my text stored?
How many tokens is an image in Gemini?
Which counts on this page are exact?
Is my text sent anywhere?
Count by model family
- Claude token counter
- GPT token counter
- Llama token counter
- DeepSeek token counter
- Mistral token counter
Or compare every family at once on the main token counter.
More from the workbench
Fine print
Tools: on the house.
Engineer: by the hour.
Counting tokens was free. Spending fewer of them is a project.
I'm Himanshu, an AI and full-stack engineer. I build MVPs, AI agents and the growth plumbing around them for early-stage startups.