# JSON module benchmarking

**URL:** <https://forum.elixirforum.com/t/json-module-benchmarking/76551>\
**Category:** Thoughts On...\
**Tags:** json\
**Created:** [September 3, 2026, 3:25pm UTC](https://forum.elixirforum.com/t/json-module-benchmarking/76551 "2026-09-03T15:25:44Z")\
**Posts on this page:** 1\
**Showing post:** 8

<div class="post-metadata">

**Author:** ![code-anth](https://forum.elixirforum.com/user_avatar/forum.elixirforum.com/code-anth/32/38238_2.png) [@code-anth](https://forum.elixirforum.com/u/code-anth)\
**Post date:** [September 4, 2026, 2:39pm UTC](https://forum.elixirforum.com/t/json-module-benchmarking/76551/8 "2026-09-04T14:39:15Z")

</div>

> [@garazdawi](#):
>
> The NIF api does not allow the user to parse erlang terms in the same speed as the JIT does, so Erlang/Elixir code written to parse erlang terms and emit binaries is faster. However, when creating Erlang terms NIFs and the JIT are about the same, so here the C code advantage wins.

Ah, thank you! That explains it. I was hoping it was something that could be optimised in the JSON, but it doesn’t sound like it. Thanks!

---

_[View the full topic](https://forum.elixirforum.com/t/json-module-benchmarking/76551)._