Every uf command timed on one generated application, beside the tool a project
would otherwise use for the same job, on one machine, in one run. Every number
below is read from the file that run wrote; none is typed in by hand. The rows
uf loses come first.
See where uf stands on each command, which
gaps are the largest, and reproduce any number on this page with one command.
Nothing has to be installed. To reproduce the numbers: a checkout of this
repository with npm ci run, and Bun on PATH for the Bun rows.
Where the numbers come from
One command measured all of them:
$ npm ci --prefix tools/bench/toolchain/rivals # the comparison tools, pinned$ uf run bench:toolchain -- --tools all --runs 7 --warmup 1
uf run bench:toolchain builds uf in release mode, writes the application
under the system's temporary directory, times every stage, prints a table and
writes a JSON file. tools/bench/toolchain/report.js documents that file. The
run below is one the Bench workflow made on a GitHub Actions runner, committed
as tools/bench/toolchain/baseline.json; this page renders that file when the
site is built.
A hosted runner is shared with other work, and a millisecond on it is not a
millisecond on your laptop. Read the ratios between rows of the same table, which
were measured minutes apart on the same machine, rather than the absolute
numbers. The load average before and after the run is below so you can see how
busy the machine was.
18 direct dependencies of a scaffolded application, pinned; `uf install` ran npm
Tool or input
Version
uf
0.0.0-alpha.46
commit
982f7923c7696c44945f5b0bccbf19457cb475e3
binary
target/release/uf
node
v24.13.0
npm
11.6.2
vite
8.2.2
react
19.3.0
host
node
vp
1.0.0-rc.0
next
16.3.6
vitest
5.0.1
bun
1.3.14
eslint
10.11.0
prettier
3.9.8
biome
2.5.12
flow
0.332.0
tsc
6.0.3
pnpm
10.34.5
Where uf loses
Every stage in which another tool's median is lower than uf's. The table is
computed from the file, so it changes when the numbers do.
Stage
uf
Faster
lint and type check, cold — small
uf check 659 ms
flow full-check 193 ms; tsc --noEmit 600 ms
test suite, cold — small
uf test 284 ms
bun test 27 ms
test suite, warm — small
uf test 74 ms
bun test 26 ms
production build, cold — small
uf build 2.55 s
vp build 296 ms
production build, warm — small
uf build 2.42 s
vp build 306 ms; next build 1.40 s
dev server start to first document, cold — small
uf dev 1.10 s
vp dev 241 ms; next dev 1.01 s
dev server start to first document, warm — small
uf dev 875 ms
vp dev 233 ms
edit to HMR update message — small
uf dev; edit components/HotCounter.js 2.4 ms
vp dev; edit src/components/HotCounter.tsx 0.9 ms
edit to updated module served — small
uf dev; edit components/HotCounter.js 9.0 ms
vp dev; edit src/components/HotCounter.tsx 2.5 ms
many small test files, cold — suite
uf test 446 ms
bun test 57 ms
many small test files, warm — suite
uf test 103 ms
bun test 60 ms
install, cold — install
uf install 7.28 s
bun install 478 ms; pnpm install 1.34 s
install, warm — install
uf install 906 ms
bun install 48 ms; pnpm install 411 ms
What each tool is given
Each tool runs on its own copy of the same application, written in the form a
project using that tool would have. tools/bench/toolchain/rivals.js generates
the copies; they share the preset's size, and the same modules and tests with
the same values and assertions.
App Router: app/layout.tsx, app/rNNN/page.tsx, the toggles marked "use client"
next dev, next build
Vitest, Rstest, Bun
the library modules and their tests against vitest, @rstest/core and bun:test
vitest run, rstest run, bun test
one copy per formatter
the Vite copy, formatted by that formatter first
vp fmt --check, prettier --check ., biome format .
Flow
the uf application with a .flowconfig, mixed spelled unknown, and flow-typed stubs for @uniflowed/*
flow full-check
install
the pinned third-party dependencies of a scaffolded application
uf install, pnpm install, bun install
many small test files
50 files × 20 cases × 2 assertions, from tools/bench/testing/suite.js
uf test, bun test, vitest run, rstest run
Where a row is not the same work, it says so here rather than in a footnote:
vp dev's first document is an empty HTML shell. Vite serves
index.html and the browser requests the application afterwards; uf and
Next.js render the page on the server before answering. The row times what a
person waits for before the first byte, and it is not the same amount of work.
next dev has no HMR rows. Next.js pushes updates over a socket protocol
of its own, and the harness speaks Vite's. vp dev's HMR is measured the way
uf's is: from the file write to the update message, and to the edited module
being served.
flow full-check sees the @uniflowed/* packages as any, through
stubs, the way a package with no library definition is seen. uf check
checks against their real types and runs uf's linter first, so it does more
work. flow full-check is the whole-project check without a server, which is
what flow check was before it became a question put to a server.
tsc --noEmit is not in the issue's table, and is here because it is
what the TypeScript copies would be type-checked with.
tsgo --noEmit is TypeScript 7's native compiler, the Go port, pinned
as typescript@7 and run as the executable its tsc launcher starts. The
launcher is a Node script that only execves that executable, so timing it
would charge TypeScript for a Node start-up it does not need. tsc stays on
TypeScript 6, the JavaScript compiler, beside it.
ESLint runs @eslint/js and typescript-eslint's recommended rules, the
configuration npm init @eslint/config writes. Biome and Vite+ run with no
configuration of their own.
Cold removes a tool's own caches before every run: .uf, dist and
node_modules/.vite for uf, .next for Next.js, node_modules/.vite for Vite+
and Vitest, node_modules/.cache for Rstest. Prettier, ESLint, Biome, tsc
and tsgo keep no cache unless asked to,
so their cold and warm rows are the same command run twice. Warm keeps
what the previous run left behind.
CPU is the user and system time of the command and every process it
waited for. It is there because a runner that spreads the same work over more
cores finishes sooner but costs the same, and on a busy machine CPU is what
runs out. A dev server's CPU is not measured, because it has not exited when
the document is served. The kernel counts CPU time in clock ticks, 10 ms on the
Linux runner, so a CPU figure under about 20 ms is a count of ticks rather
than a measurement.
Installs use the npm registry. They are measured over the network, with
an empty home directory for the cold rows, so they depend on the registry.
Every row
A table per stage, fastest first. Against uf is the ratio of medians.
format check, cold — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
uf fmt --check
0.0.0-alpha.46
13 ms
11 ms–15 ms
0.0 ms
removes .uf, dist and node_modules/.vite before every run
biome format .
2.5.12
28 ms
26 ms–30 ms
80 ms
2.1× uf
keeps no cache by default, so cold and warm are the same run
prettier --check .
3.9.8
344 ms
338 ms–347 ms
670 ms
26.7× uf
keeps no cache by default, so cold and warm are the same run
vp fmt --check
1.0.0-rc.0
422 ms
419 ms–451 ms
1.19 s
32.7× uf
keeps no cache by default, so cold and warm are the same run
format check, warm — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
uf fmt --check
0.0.0-alpha.46
12 ms
11 ms–13 ms
0.0 ms
keeps what the run before it wrote
biome format .
2.5.12
26 ms
26 ms–30 ms
80 ms
2.3× uf
keeps no cache by default, so cold and warm are the same run
prettier --check .
3.9.8
338 ms
323 ms–368 ms
670 ms
29.1× uf
keeps no cache by default, so cold and warm are the same run
vp fmt --check
1.0.0-rc.0
433 ms
416 ms–440 ms
1.19 s
37.3× uf
keeps no cache by default, so cold and warm are the same run
lint, cold — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
uf lint
0.0.0-alpha.46
28 ms
27 ms–31 ms
490 ms
removes .uf, dist and node_modules/.vite before every run
biome lint .
2.5.12
32 ms
31 ms–33 ms
160 ms
1.1× uf
keeps no cache by default, so cold and warm are the same run
vp lint
1.0.0-rc.0
272 ms
260 ms–281 ms
520 ms
9.7× uf
keeps no cache by default, so cold and warm are the same run
eslint .
10.11.0
550 ms
533 ms–571 ms
1.12 s
19.7× uf
keeps no cache by default, so cold and warm are the same run
lint, warm — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
uf lint
0.0.0-alpha.46
28 ms
25 ms–30 ms
480 ms
keeps what the run before it wrote
biome lint .
2.5.12
32 ms
31 ms–32 ms
160 ms
1.1× uf
keeps no cache by default, so cold and warm are the same run
vp lint
1.0.0-rc.0
268 ms
253 ms–274 ms
500 ms
9.4× uf
keeps no cache by default, so cold and warm are the same run
eslint .
10.11.0
550 ms
518 ms–569 ms
1.14 s
19.4× uf
keeps no cache by default, so cold and warm are the same run
lint and type check, cold — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
flow full-check
0.332.0
193 ms
175 ms–200 ms
1.49 s
3.4× faster than uf
keeps no cache by default, so cold and warm are the same run
tsc --noEmit
6.0.3
600 ms
568 ms–611 ms
1.22 s
1.1× faster than uf
keeps no cache by default, so cold and warm are the same run
uf check
0.0.0-alpha.46
659 ms
621 ms–677 ms
1.19 s
removes .uf, dist and node_modules/.vite before every run
lint and type check, warm — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
uf check
0.0.0-alpha.46
106 ms
102 ms–109 ms
650 ms
keeps what the run before it wrote
flow full-check
0.332.0
184 ms
159 ms–202 ms
1.42 s
1.7× uf
keeps no cache by default, so cold and warm are the same run
tsc --noEmit
6.0.3
590 ms
567 ms–632 ms
1.21 s
5.6× uf
keeps no cache by default, so cold and warm are the same run
test suite, cold — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
bun test
1.3.14
27 ms
25 ms–28 ms
30 ms
10.6× faster than uf
keeps no cache by default, so cold and warm are the same run
uf test
0.0.0-alpha.46
284 ms
281 ms–318 ms
2.89 s
removes .uf, dist and node_modules/.vite before every run
vitest run
5.0.1
333 ms
324 ms–343 ms
2.27 s
1.2× uf
removes node_modules/.vite before every run
vp test
1.0.0-rc.0
462 ms
455 ms–467 ms
2.46 s
1.6× uf
removes node_modules/.vite before every run
test suite, warm — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
bun test
1.3.14
26 ms
25 ms–27 ms
30 ms
2.8× faster than uf
keeps no cache by default, so cold and warm are the same run
uf test
0.0.0-alpha.46
74 ms
67 ms–77 ms
90 ms
keeps what the run before it wrote
vitest run
5.0.1
339 ms
330 ms–343 ms
2.27 s
4.6× uf
keeps what the run before it wrote
vp test
1.0.0-rc.0
467 ms
458 ms–482 ms
2.48 s
6.3× uf
keeps what the run before it wrote
production build, cold — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
vp build
1.0.0-rc.0
296 ms
288 ms–307 ms
750 ms
8.6× faster than uf
removes dist and node_modules/.vite before every run
uf build
0.0.0-alpha.46
2.55 s
2.50 s–2.56 s
5.43 s
removes .uf, dist and node_modules/.vite before every run
next build
16.3.6
3.05 s
3.02 s–3.17 s
15.56 s
1.2× uf
removes .next before every run
production build, warm — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
vp build
1.0.0-rc.0
306 ms
292 ms–322 ms
780 ms
7.9× faster than uf
keeps what the run before it wrote
next build
16.3.6
1.40 s
1.37 s–1.40 s
2.75 s
1.7× faster than uf
keeps what the run before it wrote
uf build
0.0.0-alpha.46
2.42 s
2.41 s–2.44 s
5.43 s
keeps what the run before it wrote
dev server start to first document, cold — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
vp dev
1.0.0-rc.0
241 ms
229 ms–244 ms
—
4.6× faster than uf
removes node_modules/.vite before every run
next dev
16.3.6
1.01 s
1.00 s–1.04 s
—
1.1× faster than uf
removes .next before every run
uf dev
0.0.0-alpha.46
1.10 s
1.08 s–1.12 s
—
removes .uf, dist and node_modules/.vite before every run
dev server start to first document, warm — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
vp dev
1.0.0-rc.0
233 ms
230 ms–240 ms
—
3.7× faster than uf
keeps what the run before it wrote
uf dev
0.0.0-alpha.46
875 ms
839 ms–914 ms
—
keeps what the run before it wrote
next dev
16.3.6
1.02 s
1.01 s–1.03 s
—
1.2× uf
keeps what the run before it wrote
edit to HMR update message — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
vp dev; edit src/components/HotCounter.tsx
1.0.0-rc.0
0.9 ms
0.7 ms–1.1 ms
—
2.7× faster than uf
one warm dev server; each edit rewrites one string literal
uf dev; edit components/HotCounter.js
0.0.0-alpha.46
2.4 ms
1.9 ms–9.8 ms
—
one warm dev server; each edit rewrites one string literal
edit to updated module served — small
Command
Version
Median
Min–max
CPU
Against uf
Between runs
vp dev; edit src/components/HotCounter.tsx
1.0.0-rc.0
2.5 ms
2.0 ms–3.2 ms
—
3.6× faster than uf
one warm dev server; each edit rewrites one string literal
uf dev; edit components/HotCounter.js
0.0.0-alpha.46
9.0 ms
7.7 ms–17 ms
—
one warm dev server; each edit rewrites one string literal
many small test files, cold — suite
Command
Version
Median
Min–max
CPU
Against uf
Between runs
bun test
1.3.14
57 ms
56 ms–61 ms
80 ms
7.8× faster than uf
keeps no cache by default, so cold and warm are the same run
uf test
0.0.0-alpha.46
446 ms
431 ms–487 ms
5.95 s
removes .uf, dist and node_modules/.vite before every run
vitest run
5.0.1
542 ms
535 ms–551 ms
6.32 s
1.2× uf
removes node_modules/.vite before every run
many small test files, warm — suite
Command
Version
Median
Min–max
CPU
Against uf
Between runs
bun test
1.3.14
60 ms
57 ms–61 ms
80 ms
1.7× faster than uf
keeps no cache by default, so cold and warm are the same run
uf test
0.0.0-alpha.46
103 ms
97 ms–103 ms
270 ms
keeps what the run before it wrote
vitest run
5.0.1
543 ms
539 ms–565 ms
6.30 s
5.3× uf
keeps what the run before it wrote
install, cold — install
Command
Version
Median
Min–max
CPU
Against uf
Between runs
bun install
1.3.14
478 ms
400 ms–718 ms
730 ms
15.2× faster than uf
no lockfile, no node_modules, and an empty home, so an empty package-manager cache
pnpm install
10.34.5
1.34 s
1.31 s–1.41 s
3.12 s
5.4× faster than uf
no lockfile, no node_modules, and an empty home, so an empty package-manager cache
uf install
0.0.0-alpha.46
7.28 s
7.07 s–7.38 s
7.89 s
no lockfile, no node_modules, and an empty home, so an empty package-manager cache
install, warm — install
Command
Version
Median
Min–max
CPU
Against uf
Between runs
bun install
1.3.14
48 ms
47 ms–51 ms
40 ms
18.9× faster than uf
the lockfile and package-manager cache the install before it left; no node_modules
pnpm install
10.34.5
411 ms
396 ms–419 ms
2.50 s
2.2× faster than uf
the lockfile and package-manager cache the install before it left; no node_modules
uf install
0.0.0-alpha.46
906 ms
879 ms–945 ms
1.50 s
the lockfile and package-manager cache the install before it left; no node_modules
Asked for and not measured
next hmr: next dev pushes its updates over a socket protocol of its own, and the HMR stage speaks Vite's
The regression gate
The Bench workflow also runs the uf half on its own every night, and fails
when a uf stage is more than 20% and 25 ms slower than the same row of
tools/bench/toolchain/baseline.json. Four rules keep the noise of a shared
runner from failing it:
Every row is the median of seven timed runs, after one run that is not
counted. A busy neighbour only ever slows a run down, and a median ignores
up to three slow runs out of seven.
The line is a percentage and a number of milliseconds. The format and HMR
rows take a few tens of milliseconds, and 20% of that is less than the
variation in starting a process.
A stage over the line is measured again in a new process, and the faster of
the two medians is compared. A busy neighbour rarely slows both runs; a
slower uf does.
The baseline comes from the same kind of runner the check runs on, never
from a laptop. The check warns when the CPU model differs.
Installs are measured but not gated, because the registry's speed on a given
night is not something a change to uf can affect. tools/bench/toolchain/regress.js
explains each rule.
To refresh the baseline after a change that makes uf faster or intentionally
slower, or after the runner changes: run the Bench workflow by hand, download
its bench-comparison artifact, copy its results.json over
tools/bench/toolchain/baseline.json, run uf fmt (it reformats the JSON and
changes no number), and commit it. This page is rebuilt from the new file.
What is not measured
A browser. The HMR rows stop when the edited module has been served. The
browser evaluating it and React re-rendering are not timed.
The large preset in CI.--preset large is ten times the routes and
modules and is run by hand. The nightly job runs small.
Windows. The harness times commands in process groups and reads CPU time
through a POSIX shell.
next dev HMR and vp dev server rendering, for the reasons above.
Where to go next
What uf does not do is next: what uf refuses on purpose, and
every gap with its issue number.