Token & Chunk Estimator
Paste any content to estimate tokens, how many chunks it becomes, and what embedding it will cost.
~500 tokens each with 50-token overlap.
What this estimator calculates
Paste any content — a page, a doc, an article — and this tool estimates the three numbers that decide what it costs to train a RAG assistant on it: tokens, chunks, and the embedding cost.
It mirrors a real pipeline: ~4 characters per token, ~500-token chunks with a 50-token overlap, and embedding priced like text-embedding-3-small at $0.02 per million tokens. The result is an order-of-magnitude estimate, not an invoice — but it's accurate enough to plan a build with no surprises.
Why tokens and chunks decide answer quality
Chunking is the unglamorous step that makes or breaks retrieval. Chunking 101 explains how splitting your content well keeps answers precise, and what a RAG chatbot is shows where embeddings fit in the retrieve-then-generate flow. Prefer to skip the math? Start free and let the assistant chunk, embed, and store your content for you.
Token & chunk estimator FAQ
A token is the unit language models read and bill in — roughly ¾ of a word in English, or about 4 characters. Models and embedding APIs price per token, so estimating tokens is how you estimate cost. This tool uses the same ~4-characters-per-token heuristic the indexer uses.
A chunk is a small passage your content is split into before it's embedded and stored. This estimator assumes ~500-token chunks with a 50-token overlap — a common sweet spot that keeps ideas intact without embedding whole pages as one blurry vector. The chunk count is roughly your token total divided by the step size.
Embedding is cheap. The estimate here uses text-embedding-3-small at about $0.02 per million tokens, so even a large site usually costs a few cents to embed once. Answering questions later uses a separate chat model, priced separately.
Chunks that are too big blur multiple topics into one vector and retrieve imprecisely; chunks that are too small lose the context needed to be useful. A modest overlap stops ideas from being cut mid-thought at chunk boundaries. Chunking 101 covers the trade-offs in depth.
Yes. Paste any text and it estimates tokens, chunks, and embedding cost instantly in your browser — no sign-up and nothing leaves the page.