Question 1

What is TOON format?

Accepted Answer

TOON (Token-Oriented Object Notation) is a serialization format designed to reduce the number of tokens required when sending structured data to large language models. It uses indentation-based structure, omits redundant quotes and brackets, and represents arrays of uniform objects in a tabular form — typically saving 30-60% of tokens compared to JSON.

Question 2

How much does TOON save on tokens?

Accepted Answer

On real-world workloads TOON typically saves 30-60% of tokens compared to equivalent JSON, with the largest savings on arrays of uniform objects (where it switches to a tabular format). On small deeply nested objects the savings are smaller; on highly repetitive arrays they can exceed 60%.

Question 3

Is my JSON data sent to a server?

Accepted Answer

No. Conversion happens entirely in your browser using JavaScript. Nothing is uploaded, stored or transmitted. You can use the tool offline after the first page render.

Question 4

Will TOON round-trip back to the same JSON?

Accepted Answer

Yes. The TOON output produced by this converter is lossless — when parsed with a TOON decoder it reproduces the exact same JSON structure as the input, including nested objects, arrays, primitives and edge cases like empty containers.

Question 5

Which LLMs support TOON?

Accepted Answer

TOON works with any LLM that reads text — GPT-5, GPT-4o, Claude 4, Gemini 2.5, Llama, Mistral and local models all process TOON correctly because the format is plain text. Use it as a token-efficient drop-in replacement for JSON in prompts.