Calculate how much you can save on LLM API costs using TOON format
Arrays of objects with same keys use a header row instead of repeating keys for every item.
No quotes around keys, no colons between key-value pairs, minimal punctuation.
Inline arrays and objects when they fit on one line, reducing whitespace tokens.
Yes! GPT-4, Claude, Gemini, and other modern LLMs can understand TOON format when given a brief explanation in the system prompt.
Arrays of objects with consistent keys see the biggest savings (40-60%). Simple objects save 20-30%. Deeply nested structures with varying keys save less.
Yes! TOON is a lossless format. Our converter supports bidirectional conversion with no data loss.