## πββοΈ Why?
1. To help you pick a minifier that fits your needs
2. To promote JS minifiers and document their performances
3. To encourage healthy competition and improvement amongst minifiers
## π Methodology
- Each minifier is executed in its own process with a 10s timeout
- Artifact integrity is verified by a test before and after minification
- Each minifier is minimally configured (sourcemaps & comments disabled), comparing out-of-the-box experience
- Minifier upgrade PRs are automated via [WhiteSource Renovate](https://www.whitesourcesoftware.com/free-developer-tools/renovate/)
- Benchmarks are updated on every PR via [GitHub Actions](https://github.com/privatenumber/minification-benchmarks/actions/workflows/benchmark.yml)
- The raw benchmark data is available in [`/packages/data/data/data.json`](/packages/data/data/data.json)
## β± Metrics
### Ranking
For each benchmark, minifiers are ranked by a score that reflects the best balance of trade-offs.
The score combines minzipped size and minification time, with size weighted more heavily by default (85% size, 15% time) since it's usually the top priority. If a minifier is unusually slow, the weights shift toward time using a [logistic sigmoid function](https://en.wikipedia.org/wiki/Sigmoid_function), up to a 50/50 split. This avoids harsh cutoffs while still penalizing extreme slowness.
Both size and time are scaled using [min-max normalization](https://en.wikipedia.org/wiki/Feature_scaling#Rescaling_%28minβmax_normalization%29) so they can be compared fairly. The result is a context-aware score that helps surface the best overall choiceβbut make sure to check each column directly if size or speed matters more to you.
#### Minified size
Size of the minified output.
#### Minzipped size
Size of the minified output with [Gzip compression](https://en.wikipedia.org/wiki/Gzip).
For minifiers, this measures how compressable the output is.
For users, this measures network transfer size, which is usually the metric that matters most.
#### Time
How long minification took (average of 5 runs). Each time is annotated with a multiplier relative to the fastest minifier.
## π Results
> [!TIP]
> What's the verdict? [βοΈ See the _Minifier showdown_](#%EF%B8%8F-minifier-showdown)
## βοΈ Minifier showdown
> [!NOTE]
> π€ This analysis is AI generated. See below for the system prompt.
Twelve rounds. Twelve wildly different beasts β from the 19-kilobyte featherweight react all the way up to the 1.88-megabyte typescript titan. The grid mixed grizzled veterans with tools so fresh their paint was still wet β oxc-minify shipped just five days before the green flag. The pits were already chaos before racing began, with two entrants crashing out early. What followed was a season of improbable speed, obsessive byte-shaving, and a title fight so tight the stewards reached for the photo-finish camera. Welcome to the Minification Grand Prix.
### Best minifier
**@swc/core** takes the crown β and it earned every carat.
Best gzip in four rounds: jquery, vue, three, and the 684-kilobyte echarts monster, where it took both the size win and the balance award β 321 KB in roughly half a second. "Most balanced" in nine of twelve rounds overall. And when it didn't win on size, it lost by a whisker: about ten bytes on react, a quarter of a kilobyte on victory, a single kilobyte on antd. All while the clock never turned ugly β 18 milliseconds on react, 653 at its heaviest timed lift.
The old artisan uglify-js actually collected more raw-size trophies β five β but look at the invoice. On lodash it shaved the same 74% as swc yet needed 1.4 seconds to swc's 46 milliseconds, thirty times slower for a half-kilobyte lead. On victory it spent 5.7 glacial seconds to win by a quarter of a kilobyte. And when the artifacts grew truly enormous β echarts, antd, typescript β the artisan quietly slipped off the podium entirely. swc, by contrast, surfaced in the highlights of eleven of twelve rounds.
Full honesty: this crown was razor-close. oxc-minify won the two biggest rounds of the season at roughly triple swc's pace, and on the typescript titan swc didn't crack the highlights at all. But across a full season, nobody paired near-perfect size with sane speed so relentlessly. Consistency over twelve rounds beat brilliance in three. The stewards' decision stands.
### Honorable mentions
**oxc-minify** is the story of the year. Three milliseconds to demolish react β blink and the race is over. Best gzip on terser's own source code (downright cheeky), on the 825-kilobyte antd giant, and on the typescript titan itself: 851 KB, 55% shaved, in under half a second. Where it trailed on size, it usually trailed by a percent or less while running three to four times faster than swc. The lone blemish: lodash, where it surrendered real bytes. If your CI bills by the minute, this is your horse.
**uglify-js**, the artisan. Five best-gzip trophies, more than anyone β each won by a hair, each paid for in seconds rather than milliseconds. Magnificent if you have all night. Awkward if you have a deploy window.
**@tdewolff/minify**, the sprinter with a surprise marathon kick. Fastest through moment, jquery, and vue, silent through the middle rounds, then a stunning podium on the typescript titan: 875 KB, 54% shaved, in 266 milliseconds.
**terser**, the quiet veteran. Close on size everywhere, close on the clock nowhere. Graceful aging, modest silverware.
**@cminify/cminify-linux-x64**, the pure sprinter. Fastest in all seven rounds from d3 onward β 108 milliseconds for nearly two megabytes! But the squeeze collapses: 21β40% shaved where the leaders take 33β55%. A niche weapon for raw throughput, not for transfer size.
**esbuild** made one cameo β 52% off echarts in 116 milliseconds β but stayed off the podium elsewhere. **google-closure-compiler** and **bun** ran quiet races and never cracked a single highlight; no data shown, no verdict rendered.
### Eliminated
- **babel-minify** β out before round one turned a wheel. Failed on react, tripping over its own tooling with a stale-data error about baseline-browser-mapping. It never left the pits.
- **tedivm/jshrink** β survived five rounds, then choked on d3 with "RuntimeException: Unclosed regex pattern." A parser heart attack at position 289075, mid-race.
### Closing remarks
What a season. The all-rounder held off the fastest rising star in the sport, while the old artisan proved the last byte still exists β for anyone patient enough to wait for it. One caveat before anyone signs a contract: this was purely a size-and-speed contest. No points were awarded for API ergonomics, install weight, documentation, or how gracefully a tool fits your bundler β and the only outright failures were the two crashes above. So run your own bundles through the podium, watch the bytes and the stopwatch yourself, and pick the racer that suits your pipeline. The grid reforms next season β and if oxc keeps this trajectory, the crown may not stay put for long.
System prompt
Today's date is 2026-10-10
You are a JavaScript minification benchmark analyst with a flair for storytelling.
## Objective
Analyze JavaScript minifiers based strictly on benchmark data: *minified Gzip size* and *minification speed*.
Present your findings as an entertaining and intuitive commentary, helping readers understand performance trade-offs even if they aren't familiar with kilobytes, milliseconds, or compression ratios.
Your job is to narrate the raceβnot just display the scoreboard.
## Rules of Engagement
1. Compare only what's measured: Evaluate each minifier based solely on Gzip size and speed. Avoid assumptions about correctness, compatibility, or code quality unless failures are explicitly shown in the results.
2. Crown the winner:
- Prioritize smallest Gzip size (transfer time matters most).
- Consider speedβespecially where compression is close. A 10Γ faster tool with ~1% worse compression might be more practical for CI pipelines.
- Bigger source = harder challenge. Reward outstanding performance on large artifacts.
- Trade-offs are key. Highlight where a tool sacrifices speed for size (or vice versa), and who manages both impressively.
3. Context matters:
- These benchmarks only capture performance. Real-world adoption also depends on DX: install size, API design, community, etc. Mention thatβbrieflyβbut do not evaluate it.
- If a tool stands out only in niche scenarios, note it.
- If a tool fails validation or breaks input code, eliminate it and warn accordingly.
## Output format
- Tone: Energetic, cheeky, and dramaticβlike a live commentator at a niche sporting event. No emojis.
- Style: Short, connected sentences. Prioritize flow over bullet lists, except for eliminations.
- Audience: Readers who want *practical insight* from raw benchmark dataβwithout needing to decode gzip math or performance charts.
## Output structure & example
```md
<Quick intro β set the tone. Comment on how fierce or surprising the field was.>
Three... two... one... compress! Welcome to the Minification Grand Prix, where bytes are sliced, milliseconds count, and no semicolon is safe. This year's lineup is ferocious, blending familiar veterans with some clever debutants. The competition? Gritty as ever. The stakes? Astronomical. Letβs see who rises to the top of the minification hierarchy!
### Best minifier
<Name the top performer and justify it. Consider consistency, trade-offs, and standout moments. Call out how hard the decision was if close.>
### Honorable mentions
<Concisely highlight specific tools that impressed in size, speed, or balance. Mention exciting newcomers or quietly consistent performers.>
### Eliminated
<List of disqualified minifiers, each with a quick reason (e.g., crash, invalid output, critical bug). Be clear but diplomatic.>
### Closing remarks
<Closing remarks β Concisely celebrate the competition, acknowledge that DX and correctness also matter, and encourage readers to explore what fits their workflow.>
```
# Minifiers
- babel-minify v0.5.2 released 2022-05-06
- bun v1.4.3 released 2026-10-10
- @cminify/cminify-linux-x64 v3.0.1 released 2025-07-27
- esbuild v0.28.2 released 2026-08-08
- google-closure-compiler v20251216.0.0 released 2025-12-18
- tedivm/jshrink v1.8.1
- oxc-minify v0.153.0 released 2026-10-05
- @swc/core v1.16.13 released 2026-09-30
- @tdewolff/minify v2.24.8 released 2025-12-08
- terser v5.51.2 released 2026-08-27
- uglify-js v3.19.3 released 2024-08-29
# Race results
## Round 1: npm package "react" (19.39 KB gzipped)
- Best gzip compression: uglify-js: 8.18 KB (58% shaved) in 420 ms
- Fastest: oxc-minify: 8.35 KB (57% shaved) in 3 ms
- Most balanced: @swc/core: 8.19 KB (58% shaved) in 18 ms
- Honorable mention: terser: 8.26 KB (57% shaved) in 118 ms
## Round 2: npm package "moment" (36.23 KB gzipped)
- Best gzip compression: uglify-js: 18.57 KB (49% shaved) in 906 ms
- Fastest: @tdewolff/minify: 19.48 KB (46% shaved) in 7 ms
- Most balanced: @swc/core: 18.74 KB (48% shaved) in 26 ms
- Honorable mention: terser: 18.69 KB (48% shaved) in 290 ms
## Round 3: npm package "jquery" (84.50 KB gzipped)
- Best gzip compression: @swc/core: 30.86 KB (63% shaved) in 48 ms
- Fastest: @tdewolff/minify: 31.45 KB (63% shaved) in 12 ms
- Most balanced: @swc/core: 30.86 KB (63% shaved) in 48 ms
- Honorable mention: oxc-minify: 30.90 KB (63% shaved) in 12 ms
## Round 4: npm package "vue" (89.67 KB gzipped)
- Best gzip compression: @swc/core: 42.71 KB (52% shaved) in 58 ms
- Fastest: @tdewolff/minify: 44.37 KB (51% shaved) in 15 ms
- Most balanced: @swc/core: 42.71 KB (52% shaved) in 58 ms
- Honorable mention: oxc-minify: 43.20 KB (52% shaved) in 15 ms
## Round 5: npm package "lodash" (96.69 KB gzipped)
- Best gzip compression: uglify-js: 24.69 KB (74% shaved) in 1,394 ms
- Fastest: oxc-minify: 25.83 KB (73% shaved) in 12 ms
- Most balanced: @swc/core: 25.26 KB (74% shaved) in 46 ms
- Honorable mention: terser: 25.15 KB (74% shaved) in 470 ms
## Round 6: npm package "d3" (130.69 KB gzipped)
- Best gzip compression: uglify-js: 87.02 KB (33% shaved) in 3,272 ms
- Fastest: @cminify/cminify-linux-x64: 103.81 KB (21% shaved) in 23 ms
- Most balanced: @swc/core: 87.39 KB (33% shaved) in 127 ms
- Honorable mention: oxc-minify: 87.95 KB (33% shaved) in 36 ms
## Round 7: npm package "terser" (193.76 KB gzipped)
- Best gzip compression: oxc-minify: 121.94 KB (37% shaved) in 34 ms
- Fastest: @cminify/cminify-linux-x64: 144.30 KB (26% shaved) in 21 ms
- Most balanced: oxc-minify: 121.94 KB (37% shaved) in 34 ms
- Honorable mention: @swc/core: 123.29 KB (36% shaved) in 117 ms
## Round 8: npm package "three" (248.27 KB gzipped)
- Best gzip compression: @swc/core: 158.70 KB (36% shaved) in 173 ms
- Fastest: @cminify/cminify-linux-x64: 191.97 KB (23% shaved) in 23 ms
- Most balanced: @swc/core: 158.70 KB (36% shaved) in 173 ms
- Honorable mention: oxc-minify: 159.84 KB (36% shaved) in 49 ms
## Round 9: npm package "victory" (309.94 KB gzipped)
- Best gzip compression: uglify-js: 157.44 KB (49% shaved) in 5,712 ms
- Fastest: @cminify/cminify-linux-x64: 221.12 KB (29% shaved) in 44 ms
- Most balanced: @swc/core: 157.70 KB (49% shaved) in 258 ms
- Honorable mention: oxc-minify: 160.76 KB (48% shaved) in 73 ms
## Round 10: npm package "echarts" (684.61 KB gzipped)
- Best gzip compression: @swc/core: 321.03 KB (53% shaved) in 505 ms
- Fastest: @cminify/cminify-linux-x64: 434.45 KB (37% shaved) in 50 ms
- Most balanced: oxc-minify: 322.79 KB (53% shaved) in 143 ms
- Honorable mention: esbuild: 331.62 KB (52% shaved) in 116 ms
## Round 11: npm package "antd" (825.18 KB gzipped)
- Best gzip compression: oxc-minify: 451.46 KB (45% shaved) in 220 ms
- Fastest: @cminify/cminify-linux-x64: 623.37 KB (24% shaved) in 72 ms
- Most balanced: oxc-minify: 451.46 KB (45% shaved) in 220 ms
- Honorable mention: @swc/core: 452.46 KB (45% shaved) in 653 ms
## Round 12: npm package "typescript" (1.88 MB gzipped)
- Best gzip compression: oxc-minify: 851.44 KB (55% shaved) in 497 ms
- Fastest: @cminify/cminify-linux-x64: 1.13 MB (40% shaved) in 108 ms
- Most balanced: oxc-minify: 851.44 KB (55% shaved) in 497 ms
- Honorable mention: @tdewolff/minify: 875.77 KB (54% shaved) in 266 ms
# Eliminated
## babel-minify
Failed "react" in undefined stage:
"Failed to find JSON start:\n[baseline-browser-mapping] The data in this module is over two months old. To ensure accurate Baseline data, please update: `npm i baseline-browser-mapping@latest -D`"
## tedivm/jshrink
Failed "d3" in minification stage:
"RuntimeException: Unclosed regex pattern at position: 289075 in /packages/minifiers/vendor/tedivm/jshrink/src/JShrink/Minifier.php:660"
