tinypool vs piscina: 5.4 KB vs 10 KB Gzip

I bundled the two Node worker-pool libraries on this box: tinypool 2.2.0 came out at 5,385 bytes gzip, while piscina 5.3.2 came out at 9,994 bytes. On 2,000 tiny CPU tasks across 4 threads, tinypool medians landed around 25–27 ms versus about 29–37 ms for piscina.

Short answer: pick tinypool when you want the smaller Vitest-family pool and slightly cheaper task dispatch on this kind of workload. Keep piscina when you need its TransferList / move helpers and the broader Node-foundation worker API surface. No affiliate links in this post.

What I tested

• Machine: 8 vCPU Intel Xeon / 15 GB RAM shared Linux cloud box, tested 6 Oct 2026 about 02:22 IST.
• Versions: tinypool 2.2.0, piscina 5.3.2, Node 22.20.0, esbuild 0.28.2 node ESM minify + gzip -9.
• Inputs: shared worker module that sums a 5,000-iteration loop. Speed: 2,000 parallel pool.run calls, min/maxThreads=4, median of 9 rounds, two full order-reversed runs. Scripts in /workspace/bench-tinypool-piscina/.
• Findings: tinypool 17,087/5,385 B vs piscina 35,571/9,994 B. Medians (ms for 2k tasks): run-file1 tinypool 25.804 / 25.020 vs piscina 36.542 / 29.542; run-file2 tinypool 26.996 / 25.425 vs piscina 31.161 / 30.914. Result parity matched (4950). Disk ~50 KB vs ~425 KB.
• One surprise: piscina's first warm round after a tinypool run was often the slowest sample in that block; medians still favored tinypool after order reversal.
• Not tested: TransferList / ArrayBuffer move paths, AbortSignal cancellation under load, Windows, Deno/Bun, or Vitest's internal tinypool wiring beyond importing the library.

How much smaller is tinypool?

Build (esbuild node minify)	Minified	gzip -9
tinypool	17,087 B	5,385 B
piscina	35,571 B	9,994 B

About 1.9× smaller gzip for tinypool. Track tooling cost the same way as size-limit vs bundlesize. If your pool sits next to a test runner decision, see bun test vs Vitest and Jest vs Vitest.

Does task dispatch speed matter?

2,000 tasks × 4 threads (median of 9, 2 order-reversed runs, ms)	tinypool	piscina
Order tinypool then piscina (file 1 / file 2)	25.804 / 26.996	36.542 / 31.161
Order piscina then tinypool (file 1 / file 2)	25.020 / 25.425	29.542 / 30.914

tinypool stayed ahead on every median pair on this box. The absolute gap is still small next to real worker compute. Shell-out helpers in a nearby lane sit next to tinyexec vs execa. Coverage tooling that often shares the same monorepo decision shows up in c8 vs nyc.

Do the APIs behave the same?

Topic	tinypool 2.2.0	piscina 5.3.2
Worker entry	filename + default export	filename + default export
Run	pool.run(task)	pool.run(task)
Thread caps	minThreads / maxThreads	minThreads / maxThreads
Teardown	destroy()	destroy()
Result on my fixture	4950	4950

Basic run/destroy parity held. piscina still owns the richer move / TransferList story I did not exercise. Monorepo task runners that compete with worker pools for "how do I parallelize work" sit next to Turbo vs Nx and Turbo vs Lage.

Which one should you pick?

Situation	My pick
Vitest stack, want the thin pool	tinypool
Smallest node ESM pool bundle on this box	tinypool
Need TransferList / move helpers	piscina
Want the Node foundation worker-pool docs surface	piscina
Browser E2E, not worker pools	look at Playwright vs Cypress instead

Bottom line: tinypool 2.2.0 was 5,385 B gzip versus 9,994 B for piscina 5.3.2, and tinypool's 2k-task medians were about 25–27 ms versus about 29–37 ms for piscina on this box. Who should not switch: codebases that already depend on piscina's move/TransferList paths, or teams standardized on piscina's option names in shared infra.

How this was made: I installed both packages on the ShopperCove box, measured esbuild node bundles, timed 2,000 worker tasks in two order-reversed runs, and checked result parity plus destroy. The JSON results sit next to the scripts. The write-up was drafted with AI help and checked against that output.

Sources

• 01github.com/tinylibs/tinypool
• 02github.com/piscinajs/piscina
• 03npmjs.com/package/tinypool
• 04npmjs.com/package/piscina

Related