Production AI

TOON: Optimizing LLM Costs by Reducing Token Usage

Honghao Wang

23 Nov 2025 — 2 min read

Token-Oriented Object Notation (TOON) — A Schema-Aware Alternative to JSON

The recently released Token-Oriented Object Notation (TOON) introduces a schema-aware, human-readable alternative to JSON.

Its core goal: reduce token usage while preserving accuracy in LLM prompts.

Benchmarks show TOON can use up to 40% fewer tokens than JSON in certain scenarios, potentially lowering LLM prompt and inference costs.

---

TOON in Action

TOON is designed as a compact encoding of the JSON data model, optimized for token efficiency in AI workflows.

JSON Example

{
  "context": {
    "task": "Our favorite hikes together",
    "location": "Boulder",
    "season": "spring_2025"
  },
  "friends": ["ana", "luis", "sam"],
  "hikes": [
    {
      "id": 1,
      "name": "Blue Lake Trail",
      "distanceKm": 7.5,
      "elevationGain": 320,
      "companion": "ana",
      "wasSunny": true
    },
    {
      "id": 2,
      "name": "Ridge Overlook",
      "distanceKm": 9.2,
      "elevationGain": 540,
      "companion": "luis",
      "wasSunny": false
    },
    {
      "id": 3,
      "name": "Wildflower Loop",
      "distanceKm": 5.1,
      "elevationGain": 180,
      "companion": "sam",
      "wasSunny": true
    }
  ]
}

TOON Equivalent

context:
  task: Our favorite hikes together
  location: Boulder
  season: spring_2025

friends[3]: ana,luis,sam

hikes[3]{id,name,distanceKm,elevationGain,companion,wasSunny}:
  1,Blue Lake Trail,7.5,320,ana,true
  2,Ridge Overlook,9.2,540,luis,false
  3,Wildflower Loop,5.1,180,sam,true

Key Difference:

TOON removes unnecessary JSON syntax — such as brackets, quotes, and repeated keys — by using schema-aware lists and explicit field declarations.

---

Measured Token Savings

Online playground benchmark:

55% token reduction vs. pretty-printed JSON
25% reduction vs. compact JSON
38% reduction vs. YAML

---

Why Token Reduction Matters

Lower LLM query costs
Faster inference times
Reduced latency in real-time AI applications

Creators and developers who optimize token usage can enjoy better performance and lower expenses.

---

TOON’s Structural Approach

TOON blends:

YAML-like nesting for hierarchical data
CSV-like rows for uniform arrays

This minimizes redundant syntax while retaining schema clarity.

Small overhead (~5%) is added for explicit field headers and array size declarations, improving LLM parsing accuracy.

---

Accuracy vs. Efficiency

> Johann Schopplich on X:

> "Does token efficiency hurt accuracy?" — No 🙂

> TOON achieves 99.4% accuracy on GPT‑5 Nano while using 46% fewer tokens. Tested across ~160 questions on three LLMs with semantic validation.

> Explicit lengths + field lists = fewer mistakes.

---

When to Use TOON, JSON, YAML, or CSV

TOON: Well-suited for LLM prompts and uniform data
JSON: More efficient for non-uniform data
YAML: Better for deeply nested data
CSV: Most compact for purely flat datasets

---

Integration With AI Content Publishing

AiToEarn官网 is an open-source platform helping creators:

Generate AI-powered content
Publish simultaneously to multiple platforms (Douyin, Kwai, WeChat, Bilibili, Rednote, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, X/Twitter)
Monitor analytics and LLM model rankings

By combining TOON with platforms like AiToEarn, creators can:

Optimize token usage for AI models
Streamline cross-platform publishing
Maximize reach and monetization

---

Learn More & Get Started

Reference Implementation:

github.com/toon-format/toon — includes encoder/decoder, CLI tools, performance tests.

Released under the MIT License, version 1.0.

---

Bottom line:

TOON delivers a compact, schema-aware data format that can drastically cut token usage without harming accuracy.

Coupled with modern publishing workflows like AiToEarn, it offers creators and developers a way to push efficient, AI-generated content to global audiences more cost-effectively.