Comparison

Sprae, Alpine and petite-vue do the same job: add reactivity to HTML you already have, no build step. The differences are size, speed, CSP support and maintenance. Every number below is measured — methodology at the bottom, reproduce it yourself.

At a glance

	sprae	Alpine	petite-vue
CDN build, min+gzip	11.9kb	19.9kb	7.1kb
CPU speed, geometric mean	2.81× faster	baseline	not benchmarked
Runtime memory (1k rows)	3.65MB	16.56MB	not benchmarked
First paint (1k rows)	50.6ms	71.1ms	not benchmarked
Strict CSP / no-eval	full JS expressions	restricted subset	none
Last release	active	active	Jan 2022
Reactivity	pluggable signals (TC39-track)	bundled @vue/reactivity	bundled @vue/reactivity
Keyed lists	automatic (by identity)	manual `:key`	manual `:key`
Debounce, throttle, key filters	built-in modifiers	plugin packages	none
Dependencies	0	0 (Vue reactivity inlined)	0 (Vue reactivity inlined)
Status	active, 0 open issues	active	frozen — last release Jan 2022
License	MIT	MIT	MIT

Sizes measured 2026-09-08 (sprae 13.9.2, Alpine 3.17.2, petite-vue 0.4.1); CPU speed, memory and first paint come from the official benchmark’s Chrome 152 run at sprae 13.9.1 — see Methodology.

Performance

Median times in ms (lower is better) from the official js-framework-benchmark published run, keyed suite, Chrome 152. Sprae v13.9.1 was measured in that run, on the same machine as every framework compared here — so these are directly comparable rather than stitched from separate runs:

benchmark	sprae	Alpine	ratio
create 1,000 rows	24.0	59.6	2.48×
replace all rows	26.3	71.5	2.72×
partial update (every 10th)	10.1	15.2	1.50×
select row	3.4	32.2	9.47×
swap rows	12.5	25.4	2.03×
remove row	10.4	16.6	1.60×
create 10,000 rows	246.3	606.6	2.46×
append 1,000 rows	27.3	67.2	2.46×
clear rows	10.6	61.8	5.83×
geometric mean			2.81×
memory after create 1,000 rows	3.65MB	16.56MB	4.54×
first paint	50.6	71.1	1.41×
transferred size (benchmark app)	10.8kb	14.7kb	1.36×

The rest of the field

Same run, same machine. CPU is the geometric mean over the nine benchmarks above, expressed as a multiple of sprae:

	CPU	memory after 1k	first paint
React 19.2.0	1.56× slower	4.44MB	221.4ms
Lit 3.2.0	1.13× slower	2.80MB	64.4ms
Vue 3.5.39	1.10× slower	3.93MB	93.9ms
Svelte 5.42.1	1.01× slower	2.87MB	58.4ms
sprae 13.9.1	—	3.65MB	50.6ms
plain DOM (vanillajs)	1.16× faster	1.86MB	52.7ms

Read that last row honestly: hand-written DOM is still faster on CPU, by about 16%. Lit and Svelte hold less memory than sprae, and both ship a compiler to do it. First paint against plain DOM is a wash — 50.6ms against 52.7ms is inside single-run noise, so it means “no measurable cost”, not a win.

Run it independently: krausest/js-framework-benchmark includes both frameworks.

When Alpine is the better choice

Honesty over conversion:

Laravel Livewire — Alpine ships inside it; use what’s already there.
Official plugins — x-transition sugar, mask, collapse, sort, persist are packaged and documented. Sprae covers most of this with modifiers and CSS, but you assemble it yourself.
Community volume — years of StackOverflow answers, recipes, and tutorials. Sprae trades ecosystem size for engine size.

When petite-vue is the better choice

You want exact Vue template syntax and the smallest possible script, and you accept development frozen since January 2022 (npm).

Migrating

Alpine → sprae — directive-by-directive mapping.
petite-vue → sprae — mostly rename: v-scope→:scope, v-if→:if, v-for→:each, @click→:onclick, v-model→:value.

Methodology

Sizes measured 2026-09-08, unpkg default build of each package (sprae 13.9.2, Alpine 3.17.2, petite-vue 0.4.1):

curl -sL https://unpkg.com/sprae      | gzip -9 | wc -c   # 11932
curl -sL https://unpkg.com/alpinejs   | gzip -9 | wc -c   # 19865
curl -sL https://unpkg.com/petite-vue | gzip -9 | wc -c   #  7061

Both grew since the July run (sprae 10947 → 11932, Alpine 16694 → 19865); Alpine grew faster, so the gap widened.

sprae’s ESM build (sprae.js) is smaller — 10.9kb gzip, 9.9kb brotli — if you import it directly.

CSP builds are compared separately on the strict CSP page: @alpinejs/csp has no unpkg field, so unpkg.com/@alpinejs/csp resolves to its CommonJS main rather than a browser bundle — compare dist/cdn.min.js on both sides or the numbers are not like-for-like.

Performance: the official js-framework-benchmark published run for Chrome 152, keyed suite, medians of 15 samples. Sprae v13.9.1 is in that run, so every framework on this page was measured on one machine in one session — nothing is stitched together from separate runs. Memory is the harness’s GC’d heap after create-1k; transferred size and first paint are measured by the harness itself.