JSON vs JSON5 vs JSONC: Differences and When to Use Each

October 30, 2026 · 4 min read

JSON is the strict data format every language can parse; JSONC is JSON plus comments (and, in most parsers, trailing commas), used for config files like VS Code settings and tsconfig.json; JSON5 goes further with single quotes, unquoted keys, hex numbers and more. Use JSON for anything exchanged between programs, and JSONC or JSON5 only for hand-written config read by a parser that explicitly supports it.

Feature comparison

The same config in each format

{
  "name": "api",
  "port": 8080,
  "retries": 3,
  "allowedOrigins": ["https://example.com"]
}

{
  // Service name used in logs
  "name": "api",
  "port": 8080,
  /* Retries for upstream calls */
  "retries": 3,
  "allowedOrigins": ["https://example.com"],
}

{
  // Service name used in logs
  name: 'api',
  port: 0x1F90, // 8080
  retries: 3,
  allowedOrigins: ['https://example.com',],
}

Where you meet each one

JSON

API requests and responses, message queues, data files, package.json, browser storage. Its strictness is the point: any compliant parser in any language reads it the same way, and there is no ambiguity about what a document means.

JSONC

VS Code's settings.json, launch.json and tasks.json, TypeScript's tsconfig.json and jsconfig.json, and Dev Container devcontainer.json. The common thread is files people edit by hand and want to annotate. Some tools that accept comments in their config still read plain JSON elsewhere: package.json, for example, must be strict JSON.

JSON5

Config for some JavaScript tooling (Babel parses .babelrc with a JSON5 parser, and Renovate supports renovate.json5), and some game and app configs. It is more comfortable to write but you need the json5 library to read it.

Parsing JSONC and JSON5 in code

// npm install json5 jsonc-parser
import JSON5 from "json5";
import { parse as parseJsonc } from "jsonc-parser";

JSON5.parse("{name: 'api', port: 0x1F90,}");   // { name: "api", port: 8080 }
parseJsonc('{ "a": 1, // note\n }');           // { a: 1 }

// Convert either to strict JSON for other tools
const strict = JSON.stringify(JSON5.parse(source), null, 2);

In Python, the standard json module only reads strict JSON; third-party packages such as json5 handle JSON5. Avoid stripping comments with a regular expression: a // inside a string such as a URL will be removed too.

Converting to strict JSON

Comments are lost when you convert, because JSON has nowhere to put them. Everything else maps cleanly, except Infinity and NaN, which have no JSON equivalent and usually become null. Our JSON formatter has a JSON5 input mode that accepts JSON5 and JSONC and shows the strict JSON it means, with a note for anything that changed. For a file with only a few stray comments or trailing commas, the JSON fixer removes them and shows a diff.

Which should you choose?

• Data exchanged between systems: JSON, always. Every consumer can read it and nobody has to guess which extensions are allowed.
• A config file your tool already defines: whatever the tool documents. Do not add comments to a file the tool reads as strict JSON.
• A new config format you are designing: JSONC is a modest, widely understood step. If you want much friendlier syntax, consider YAML or TOML, which have broader ecosystem support than JSON5.
• Data you will validate with JSON Schema: schemas apply to the parsed data, so they work with any of the three once parsed. See the JSON Schema beginner guide.

If a strict parser rejects a file you thought was JSON, run it through the validator. An error at a / or at a closing bracket after a comma is the giveaway that it is really JSONC or JSON5.