Dataset Size Calculator
Turn rows and average length into tokens and storage.
Before you embed a corpus or fine-tune on a dataset, it helps to know how big it is in tokens, because that's what you pay for. This training data size estimator does the arithmetic: enter your row count and the average length of a row in tokens, words or characters, and it returns the total token count, a raw-text storage estimate and an embedding cost. Use it as a rows to tokens calculator for a fine-tune, or an ml dataset token count check before an indexing run.
Your dataset
Estimated total tokens
33,250,000
≈ 33.25M tokens
- Raw text size
- 127 MB
- Characters
- 133,000,000
- Embed · 3-small
- $0.67
How this is estimated
Tokens come from your row count times average length, converting words at 1.33 tokens each and characters at 4 per token. Raw-text size assumes about one byte per character — a plain-text estimate, before any database index or vector storage on top.
Token counts are heuristics that vary by tokenizer and language, so treat the total as an order-of-magnitude figure for sizing a training or embedding run. Prices updated January 2026. Storing the vectors themselves takes far more than the raw text — check our embedding cost calculator for that.
How it works
- 1
Enter your rows
Put in the number of rows, examples or documents in the dataset. This is the count the average length gets multiplied across.
- 2
Set the average length
Enter a typical row's length and choose the unit — tokens if you already know them, or words or characters, which the tool converts to tokens for you.
- 3
Read size and cost
The headline is the estimated total tokens. The stats show the raw-text storage size, the character count and what it would cost to embed the whole set with the model you pick.
Instant & 100% private — nothing is uploaded
Every calculation runs locally in your browser. The prompts, token counts and numbers you enter stay on your own device and are never sent to a server — nothing is stored, logged or shared.
Frequently asked questions
- How do I convert rows to tokens?
- Multiply the number of rows by the average tokens per row. If you know your rows in words or characters instead, this tool converts them — roughly 1.33 tokens per word, or one token per four characters of English — then multiplies across your row count for the dataset total.
- How much storage will the dataset need?
- The raw text is about one byte per character, and there are roughly four characters per token, so a five-million-token dataset is around 20 MB of plain text. That's the text itself only — a database index, or the embedding vectors you generate from it, take considerably more room on top.
- Why does the token count matter for training or embedding?
- Both fine-tuning and embedding bill by the token, so the token count is what sets the cost, and it also decides whether the job fits a model's limits. Knowing it up front lets you estimate the bill and spot a dataset that's larger than you expected before you kick off a long run.
- How accurate is the estimate?
- It's an order-of-magnitude figure. Real token counts depend on the tokenizer, the language and how much punctuation, code or markup the text contains, all of which shift the words-to-tokens ratio. For a firm number, tokenize a representative sample and scale it up; use this for planning and budgeting.
Important
For planning and estimates only. Prices come from a published rate table dated on the page; providers change pricing without notice, and token counts here are approximations. Confirm against the provider’s own pricing before you budget or commit.
More tools
More from the Hivly network
Free sister tools on our other sites.
