native-recipe performance

• 5×as fast with the cache, for a style recipe
• 11×as fast with the cache, for a slot style recipe
• 95%of the speed of a style recipe, for a themed one

What makes a style cheap for React Native is its identity, not StyleSheet.create, which returns the styles it is given and only checks their types. A recipe returns the same frozen object for the same variants, so the style prop keeps its identity between renders, and caches it under one integer per selection, so most calls are a lookup (see How it works).

The cache

The benchmarks call a style recipe, and a slot style recipe, which builds a style for each slot, with five selections in turn, which cover default variants, compound variants, and boolean variants. Each compares the recipe with the same recipe built on every call.

Cache benchmark

Cache	Relative to the fastest of its group	Calls/s	Compared
Style recipe, with cache		15.7M	—
Style recipe, without cache		3.24M	4.9× slower
Slot style recipe, with cache		15.3M	—
Slot style recipe, without cache		1.44M	10.6× slower

Create each recipe once

The cache belongs to the recipe, so create each recipe at the top level of a module, not inside a component. It grows with the selections the recipe is called with, up to one entry per combination of declared options.

Themed recipes

A themed recipe finds the recipe of its theme, then calls it, still without allocations. It keeps the recipe of the last theme apart, so calling it again with the same theme skips the lookup; another theme is found in a WeakMap. The benchmark calls a themed recipe with the five selections, twice in each iteration: with one theme both times, or with a light theme, then a dark one. It compares it with the same recipe without a theme, and with the recipe of each theme built on every call.

Themed recipe benchmark

Themed recipe	Relative to the fastest of its group	Calls/s	Compared
Style recipe		15.9M	1.1× faster
Themed style recipe, one theme		15.1M	—
Themed style recipe, two themes		14.6M	on par
Style recipe, without cache		3.49M	4.3× slower

Method

Measured on

October 6, 2026

On an Apple M1 Pro with Node.js 24.21.0, against:

• native-recipe 1.1.0

• The benchmarks first check that each recipe returns the expected styles, for each theme.
• They import the package by its name, so they run the built bundle, not the sources.
• The benchmarks assert that a style recipe, a slot style recipe, and a themed recipe are faster with their cache than without it.

Run them with pnpm bench, which builds the package first, or with pnpm measure, which also saves the results shown here.