Convert JSON to TOON format to cut LLM prompt tokens by ~40%. TOON achieves higher LLM accuracy on the same data with fewer tokens. Lossless and bidirectional.
Share this tool
Found it useful? Help a fellow developer discover it.
| Format | Tokens (mixed) | LLM Accuracy | Best for |
|---|---|---|---|
| TOON | 2,744 | 73.9% | Uniform arrays of objects |
| JSON compact | 3,081 | 70.7% | General-purpose API data |
| YAML | 3,719 | 69.0% | Human-readable configs |
| JSON | 4,545 | 69.7% | Standard pretty-printed JSON |
| XML | 5,167 | 67.1% | Legacy / SOAP systems |
Benchmark data from the TOON spec across 209 retrieval questions on 4 LLM models.
TOON (Token-Oriented Object Notation) is a compact, lossless encoding of the JSON data model designed for LLM input. It combines YAML-style indentation for nested objects with CSV-style tabular layout for uniform arrays of objects. This means a 100-row array of employee records where every row has the same fields is encoded as a header line followed by comma-separated rows, rather than repeating the field names 100 times as JSON does. The result is approximately 40% fewer tokens with no loss of information. The conversion is lossless: JSON → TOON → JSON produces identical output. TOON's sweet spot is uniform arrays of objects. For deeply nested or non-uniform JSON, the tool will show an advisory.
Yes. TOON encodes the same data model as JSON: strings, numbers, booleans, null, arrays, and objects. The conversion is fully lossless, running JSON → TOON → JSON produces byte-for-byte identical output.
TOON is most beneficial for uniform arrays of objects (same fields across all rows). For deeply nested configs, non-uniform structures, or single objects with no arrays, the token savings are minimal or negative. The tool will display an advisory in those cases.
Any LLM that can process text can read TOON. No special model support is required. TOON is a text format. Benchmarks show it works well across GPT-4, Claude, Gemini, and Llama model families.
Yes. The @toon-format/toon npm package is browser-compatible and MIT licensed. Install it with npm install @toon-format/toon. This tool uses that same package for all conversions.
TOON's tabular format is only applied to uniform arrays, meaning arrays where every object has the same set of keys with primitive values. For JSON that is mostly nested objects or irregular arrays, TOON falls back to YAML-style indentation, which may not be significantly smaller than compact JSON.
Benchmarks across 4 LLM models show TOON uses approximately 40% fewer tokens than pretty-printed JSON and about 11% fewer than compact (minified) JSON when applied to uniform arrays of objects. For deeply nested or non-uniform JSON the reduction is smaller or negligible. The tool shows a live token count comparison when you convert.
Paste your JSON into the input panel and the TOON output appears instantly on the right. The tool automatically detects uniform arrays and applies tabular encoding where beneficial. Click Copy to copy the TOON output or Download to save it as a file. To convert back, toggle the direction to TOON to JSON.
TOON is an open format. The specification and reference implementation are published under an open license. The npm package @toon-format/toon is MIT licensed and browser-compatible. This tool uses that same package for all conversions, so the output is identical to what you would get programmatically in your own application.
JSON Formatter & Validator
Format, beautify, validate, and minify JSON. Detects BigInt precision issues, duplicate keys, and JSONC comments.
AI Token Counter
Count tokens for GPT-4o, GPT-4, GPT-3.5, Claude, and Gemini. See context window usage and estimated API cost.
JSON Diff
Compare two JSON objects side-by-side and highlight differences.
Online Notepad
Autosaving plain-text editor with word count, Find & Replace, and .txt download. Nothing leaves your browser.
