What chunk size should I use for RAG?

Common chunk sizes range from 256 to 1024 tokens. Smaller chunks (256-512) give more precise retrieval but may lose context. Larger chunks (512-1024) preserve more context but may include irrelevant content. Start with 512 tokens and tune based on your retrieval quality.

How much overlap should I add between chunks?

An overlap of 10-20% of your chunk size is typical. For a 1000-token chunk, 100-200 tokens of overlap helps preserve context at chunk boundaries. More overlap increases storage and computation costs.

Why split at word boundaries instead of exact token counts?

Token boundaries often fall in the middle of words. Splitting at word boundaries ensures clean, readable chunks that embed more naturally and are easier to debug.

AI Chunk Overlap Tool

Split text into token-sized chunks with configurable overlap for RAG and embedding pipelines.

Input text

Chunk size: 1000 tokens

1004000

Overlap: 100 tokens

0500

Paste text above to see chunks

Related Tools

ATCAI Text CleanerNEW

Clean and sanitize text for LLM input by stripping HTML, normalizing Unicode, and collapsing whitespace.

AIPAI Input PreprocessorNEW

Full preprocessing pipeline for LLM input: trim, normalize, strip HTML, collapse whitespace, and truncate to context window.

TKNAI Token CounterNEW

Count tokens for GPT, Claude, Gemini, and LLaMA models.

ADFAI Dataset FormatterNEW

Convert CSV, TSV, or JSON data to JSONL format for LLM fine-tuning with role mapping.

Learn More

guide:chunking strategies use case:rag pipelines

FAQ

What chunk size should I use for RAG?: Common chunk sizes range from 256 to 1024 tokens. Smaller chunks (256-512) give more precise retrieval but may lose context. Larger chunks (512-1024) preserve more context but may include irrelevant content. Start with 512 tokens and tune based on your retrieval quality.
How much overlap should I add between chunks?: An overlap of 10-20% of your chunk size is typical. For a 1000-token chunk, 100-200 tokens of overlap helps preserve context at chunk boundaries. More overlap increases storage and computation costs.
Why split at word boundaries instead of exact token counts?: Token boundaries often fall in the middle of words. Splitting at word boundaries ensures clean, readable chunks that embed more naturally and are easier to debug.

Split long documents into overlapping chunks for retrieval-augmented generation (RAG) and vector embedding. Configure chunk size (tokens) and overlap size with sliders. Chunks are split at word boundaries using a token estimator. See numbered chunk cards with token counts and overlap indicators.