# TOON

13 Hacker News commenters recommended TOON and 10 criticized it, November 2025 to November 2026. It was recommended in 8 of 13 months.

**Verdict:** People recommend it for cutting tokens when sending structured data to LLMs, since field names are declared once. Critics say savings vanish against optimized JSON.

## Why people recommend it

- Fewer tokens than JSON (3 comments): "TOON was much more compact, but when researchers examined whole-session effects, it was a wash" ([source](https://news.ycombinator.com/item?id=49253910)) "TOON's structural overhead is lower than compact JSON for the structure families analyzed here" ([source](https://news.ycombinator.com/item?id=46077959)) "TOON format: 50-58% token reduction for LLM interactions" ([source](https://news.ycombinator.com/item?id=46505215))
- Works well in practice (3 comments): "TOON is a real game changer when used as an intermediary between the IDE/chat and the LLM" ([source](https://news.ycombinator.com/item?id=46121709)) "LLMs parse it correctly in our testing (GPT-4o, Claude, etc)." ([source](https://news.ycombinator.com/item?id=46882036)) "that works remarkably well too" ([source](https://news.ycombinator.com/item?id=47267401))

## What people warn about

- Savings vanish in whole-session tests (3 comments): "TOON was much more compact, but when researchers examined whole-session effects, it was a wash" ([source](https://news.ycombinator.com/item?id=49253910)) "TOON’s advertised 40–60% savings dropped to ~2% overall once JSON was fully optimized." ([source](https://news.ycombinator.com/item?id=46094714)) "suggesting that TOON doesn't help overall" ([source](https://news.ycombinator.com/item?id=48641753))
- Models unfamiliar with the format (3 comments): "Toon is 3 months old, it's not used by anyone, and it's therefore not in the training set of any model." ([source](https://news.ycombinator.com/item?id=46654313)) "LLMs were less familiar with it, and thus spent even more tokens trying to decipher it" ([source](https://news.ycombinator.com/item?id=46999484)) "Is there evidence that LLMs adhere to this format better than to JSON? I doubt that." ([source](https://news.ycombinator.com/item?id=46653908))
- Worse than JSON in some tests (2 comments): "TOON performed worse than both JSON and having the LLM just consume the human output." ([source](https://news.ycombinator.com/item?id=48057811)) "It’s the same story we’ve seen before with TOON." ([source](https://news.ycombinator.com/item?id=46585046))

## Alternatives hackers recommend

- [JSON5](https://hackerapproved.com/p/json5.md): 15 recommend

## Cite as

Hacker Approved, "TOON", https://hackerapproved.com/p/toon/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).