Guide

Performance Test Results

Benchmarks vs nuxt-i18n on real fixtures.

Project Information

  • plain-nuxt: ./scripts/perf-fixtures/plain-nuxt
  • i18n-micro: ./scripts/perf-fixtures/i18n-micro
  • i18n: ./scripts/perf-fixtures/i18n
  • CLI: pnpm test:performance

Description

Compares plain Nuxt, i18n-micro, and @nuxtjs/i18n under one shared dictionary profile via untestutils/perf.

Focus: build time, peak RSS, deployable code vs translations (asset) vs total, and load (Artillery).

Methodology notes

  • Metrics are means of 3 consecutive runs per fixture (not interleaved).
  • Artifact: Nitro node-server .output/server/index.mjs after forced nuxi build; nuxt-i18n-micro is the built module dist (not dev:prepare jiti stub). See benchmark methodology.
  • Translations = bundle.asset (locale JSON, chunks/raw/, matching locale chunks via isTranslationFile).
  • Code excludes source maps (*.map): the fixtures link the module from the repo, so Nitro writes full maps for it and none for npm dependencies.
  • plain-nuxt serves the same leaf volume as static JSON — I/O-heavy under load, not “i18n overhead”.
  • Load profile --load full: warm 6s@6 + main 60s@60 uncapped (historical).
  • Cool-downs --cool fast (see methodology doc).
  • Each of the runs builds is forced (no warm-cache zeroing of build time).
  • Artillery Error rate in tables is HTTP-status errors from the load metrics. Charts also plot vusers.failed (client timeouts under saturation); those can be high while HTTP error rate stays 0%.

Runs

All metrics below are means across 3 runs.

Fixture profile

KnobValue
Locales4 (en, de, ru, fr)
Pages2 (index, page)
Index treedepth 5, branch 7 → 16,807 leaf keys / locale
Secondary pagesdepth 5, branch 6 → 7,776 leaf keys / page / locale
GoalDefault CLI profile (--locales 4 --keys 10000); raise knobs for regression-radar loads

Dictionaries come from scripts/perf-fixtures/perf-shared/runtime.json (written by the CLI) so all three fixtures stay aligned.

Dependency Versions

DependencyVersion
nodev22.23.4
nuxt4.6.4
nuxt-i18n-micro4.0.0
@nuxtjs/i18n10.6.0

Source dictionaries (pre-build)

FixtureOn-disk locale data
plain-nuxt8.25 MB
i18n-v106.8 MB
i18n-micro6.8 MB

From runtime.json (4 locales, 16,807 index leaf keys).

Build Performance for scripts/perf-fixtures/plain-nuxt

  • Build Time: 5.92 seconds
  • Bundle Size: 7.91 MB (code: 1.11 MB, translations: 6.8 MB)
  • Output dirs: public: 7.01 MB, server: 971.23 KB
  • Max / Avg CPU: 237.93% / 124.86%
  • Max / Avg Memory: 1042.92 MB / 447.42 MB

Build Performance for scripts/perf-fixtures/i18n

  • Build Time: 6.18 seconds
  • Bundle Size: 8.96 MB (code: 1.75 MB, translations: 7.21 MB)
  • Output dirs: public: 289.27 KB, server: 8.82 MB
  • Max / Avg CPU: 237.80% / 150.36%
  • Max / Avg Memory: 1356.10 MB / 660.55 MB

Build Performance for scripts/perf-fixtures/i18n-micro

  • Build Time: 4.74 seconds
  • Bundle Size: 8.09 MB (code: 1.29 MB, translations: 6.8 MB)
  • Output dirs: public: 7.06 MB, server: 3.1 MB
  • Max / Avg CPU: 247.27% / 158.03%
  • Max / Avg Memory: 1195.70 MB / 584.05 MB

Build Performance Summary (mean of 3 runs)

ProjectBuild TimeCode BundleTranslationsTotal
plain-nuxt (baseline)5.92s1.11 MB6.8 MB7.91 MB
i18n-v106.18s1.75 MB7.21 MB8.96 MB
i18n-micro4.74s1.29 MB6.8 MB8.09 MB

“Total” = code + translations (bundle.asset). Translations include locales/, _locales/, chunks/raw/, and matching locale chunks.

Loading chart…
Loading chart…
Loading chart…
Loading chart…

Load Results for plain-nuxt

Resource Usage

  • Max / Avg CPU: 142.00% / 118.34%
  • Max / Avg Memory: 498.61 MB / 402.79 MB

Artillery

  • Duration: 86.17s · RPS: 114.67 · Error rate: 0.00%
  • Latency avg / p50 / p95 / p99: 1744.23 / 608.33 / 11585.40 / 12474.77 ms
Loading chart…
Loading chart…

Load Results for i18n-v10

Resource Usage

  • Max / Avg CPU: 134.43% / 104.91%
  • Max / Avg Memory: 482.12 MB / 369.27 MB

Artillery

  • Duration: 77.07s · RPS: 178.00 · Error rate: 0.00%
  • Latency avg / p50 / p95 / p99: 757.87 / 86.87 / 6074.27 / 6428.17 ms
Loading chart…
Loading chart…

Load Results for i18n-micro

Resource Usage

  • Max / Avg CPU: 137.13% / 98.79%
  • Max / Avg Memory: 407.34 MB / 327.15 MB

Artillery

  • Duration: 66.13s · RPS: 414.67 · Error rate: 0.00%
  • Latency avg / p50 / p95 / p99: 41.17 / 28.60 / 101.87 / 237.73 ms
Loading chart…
Loading chart…

Load Summary (mean of 3 runs)

Artillery

ProjectAvg ResponseP50P95P99RPSError Rate
plain-nuxt1744.23 ms608.33 ms11585.40 ms12474.77 ms114.670.00%
i18n-v10757.87 ms86.87 ms6074.27 ms6428.17 ms178.000.00%
i18n-micro41.17 ms28.60 ms101.87 ms237.73 ms414.670.00%

Performance Comparison

Throughput (Requests per Second)

Winner: i18n-micro with 415 RPS

Loading chart…

Latency Distribution

Winner: i18n-micro with 41.17 ms avg latency

Loading chart…

Quick Comparison

Metricplain-nuxti18n-v10i18n-microBest
RPS (Artillery)115178415i18n-micro
Avg Latency1744.23 ms757.87 ms41.17 msi18n-micro
P99 Latency12474.77 ms6428.17 ms237.73 msi18n-micro
Error rate0.00%0.00%0.00%-

Comparison: plain-nuxt (baseline) vs i18n v10

Metricplain-nuxt (baseline)i18n v10Difference
Max Memory498.61 MB482.12 MB-16.49 MB
Avg Memory402.79 MB369.27 MB-33.52 MB
Response Avg1744.23 ms757.87 ms-986.37 ms
Response P9511585.40 ms6074.27 ms-5511.13 ms
RPS (Artillery)114.67178.00+63.33
Error rate0.00%0.00%0.00 %

Comparison: plain-nuxt (baseline) vs i18n-micro

Metricplain-nuxt (baseline)i18n-microDifference
Max Memory498.61 MB407.34 MB-91.27 MB
Avg Memory402.79 MB327.15 MB-75.64 MB
Response Avg1744.23 ms41.17 ms-1703.07 ms
Response P9511585.40 ms101.87 ms-11483.53 ms
RPS (Artillery)114.67414.67+300.00
Error rate0.00%0.00%0.00 %

Comparison: i18n v10 vs i18n-micro

Metrici18n v10i18n-microDifference
Max Memory482.12 MB407.34 MB-74.78 MB
Avg Memory369.27 MB327.15 MB-42.12 MB
Response Avg757.87 ms41.17 ms-716.70 ms
Response P956074.27 ms101.87 ms-5972.40 ms
RPS (Artillery)178.00414.67+236.67
Error rate0.00%0.00%0.00 %

Notes

  • Shared profile: 4 locales × 2 pages × ~16.8k index leaves.
  • Load: programmatic Artillery only — warm 6s@6 + main 60s@60, uncapped VU (historical YAML methodology). Paths from runtime profile — see scripts/src/perf/load.ts.
  • Cool-downs: 2s post-build, 3s between runs, 5s between fixtures. Builds are forced each run so means are not diluted by cache hits.
  • Re-run day-to-day: pnpm test:performance (--load short --cool fast).
  • Regenerate this page: pnpm -C scripts cli performance --load full --only all --runs 3.
  • Flags: --locales N --keys K --only all|micro|i18n|plain --runs N --skip-load --cool fast|strict.

Performance tuning

The settings that change memory, payload size and caching, and how to measure what your app ships.

DevTools

Edit translation files in the Nuxt DevTools tab and inspect locale state in Vue DevTools during development.