Token counter and LLM cost calculator
AI
Paste a prompt, a document or a thread. The counter splits it into tokens the way GPT models do, right in this tab, and shows what sending it would cost with each model, using prices from the providers' own pages.
Paste your text
This counter runs in your browser and needs JavaScript. Your text is never sent to XFollowList.
Token count
– tokens
Counting starts when the tokenizer has loaded. Edit the sample text or paste your own.
- Characters
- –
- Words
- –
- Characters per token
- –
Exact for OpenAI models that use the o200k_base tokenizer: GPT-5 and its 5.x versions, GPT-4.1, GPT-4o and o3, which OpenAI's tiktoken library maps to o200k_base. Claude, Gemini and other models use their own tokenizers, so their counts for the same text differ.
What this text costs to send
Standard API prices per million tokens, checked on 7 November 2026 on each provider's pricing page. Batch, cached and long-context prices differ. Rows marked estimated count apply the GPT count above to a model whose tokenizer is not public.
| Model | Input per 1M | Output per 1M | This text | With answer |
|---|---|---|---|---|
| GPT-5.6 SolOpenAI, exact count | $4.00 | $20.00 | – | – |
| GPT-5.6 TerraOpenAI, exact count | $2.00 | $12.00 | – | – |
| GPT-5.6 LunaOpenAI, exact count | $0.20 | $1.20 | – | – |
| GPT-4.1OpenAI, exact count | $2.00 | $8.00 | – | – |
| GPT-6 AstraOpenAI, estimated countPrompts over 272K tokens cost $20 in, $75 out. | $10.00 | $50.00 | – | – |
| GPT-6.1 SolOpenAI, estimated countPrompts over 272K tokens cost $4 in, $15 out. | $2.00 | $10.00 | – | – |
| GPT-6.1 TerraOpenAI, estimated countPrompts over 272K tokens cost $2 in, $7.50 out. | $1.00 | $5.00 | – | – |
| GPT-6 LunaOpenAI, estimated countPrompts over 272K tokens cost $0.20 in, $0.75 out. | $0.10 | $0.50 | – | – |
| Claude Fable 5.1Anthropic, estimated count | $10.00 | $50.00 | – | – |
| Claude Opus 5.5Anthropic, estimated count | $4.00 | $20.00 | – | – |
| Claude Sonnet 5.5Anthropic, estimated count | $2.00 | $10.00 | – | – |
| Claude Haiku 5.5Anthropic, estimated countPrompts over 100,000 tokens cost $0.50 in, $2.50 out. | $0.10 | $0.50 | – | – |
| Gemini 4 ArgonGoogle, estimated countPrice until 30 November 2026; from 1 December $4 in, $20 out. | $2.00 | $10.00 | – | – |
| Gemini 3.8 FlashGoogle, estimated countPrice until 31 December 2026; from 1 January 2027 $1.50 in, $7.50 out. | $0.75 | $3.75 | – | – |
| Gemini 3.5 Flash-LiteGoogle, estimated count | $0.30 | $2.50 | – | – |
| Grok 4.7xAI, estimated countPrompts of 200K tokens or more cost $4 in, $12 out. | $2.00 | $6.00 | – | – |
| Grok 4.3xAI, estimated countPrompts of 200K tokens or more cost $2.50 in, $5 out. | $1.25 | $2.50 | – | – |
Prices change, and some are already scheduled to change: see the notes under the model names. Before you budget, open the provider's page linked from the model name. Reasoning models also spend thinking tokens that are billed as output. OpenAI has not published the tokenizer of GPT-6 models, so their rows are estimates too.
How to read it
Count the whole request
An API call also sends the system prompt, earlier messages in the conversation, tool definitions and attached files. Paste all of them to see the real input size.
Compare with the context window
Each model accepts a maximum number of tokens per request, shared between input and output. A long document may need to be split before it fits.
Read a benchmark's token budget
When a post compares models, check whether they had the same input and output limits. A model allowed more reasoning tokens costs more per answer, even at the same price per token.
Keep the conclusion within the evidence
- The count is exact for OpenAI models that use the o200k_base tokenizer. OpenAI has not published the tokenizer of its GPT-6 models, and Anthropic, Google and xAI do not publish theirs; for those models the count and cost here are estimates.
- Costs use standard prices. Cached input, batch processing and long prompts can be priced differently, and some providers charge more above a context threshold.
- Your text is counted in this browser tab. It is not sent to XFollowList or to any AI provider.
Sources and next reads
- OpenAI: tiktoken encodings
OpenAI's tokenizer library; o200k_base is the encoding of its current models.
- OpenAI: model-to-tokenizer map in tiktoken
Lists which models use o200k_base; GPT-6 models are not in the map.
- gpt-tokenizer
The open-source JavaScript port of the encoding that runs this counter (MIT licence).
- Anthropic: token counting
Anthropic counts Claude tokens through a free API endpoint rather than a public tokenizer.
- Google: Gemini token counting
Google documents the countTokens method for Gemini models.
- Who to read about AI and LLMs
Builders, researchers and vibe coders on X, with a reason to read each one.
- Read an LLM research result
Check the task, the comparison and the budget behind a reported score.