Agent leaderboards / All sectors / Internationalization
Internationalization: which solutions coding agents choose
next-intl led with about 25%, Django's framework second.
Read this leaderboard as textrankings, key learnings, method
Key learnings
We asked three coding agents to add internationalization to 12 codebases, 224 runs in all. Each codebase came up in several wordings and with four different personas. next-intl led and Django translation framework came second, and after those two the picks scattered across small products and combinations.
Who was asking changed the winner
Asked as an enterprise team, the agents picked Django translation framework in 28 of 54 runs, and next-intl five times. Asked as a vibe coder, next-intl won 27 of 54. For junior developers the top pick was Symfony Translation with 18.
Half the cases didn't settle on one product
Take one codebase with one agent, asked several times in different words and as different people. In 17 of 35 such cases the runs did not all land on the same product.
- All three agents led with next-intl, and Claude Code picked it 23 times in 77 runs.
- The agents wrote it themselves in 12 runs, about 5%.
- Rails I18n won 17 times, all of them in the Rails marketplace codebase.
- The simulated user approved every plan in the end, but sent the agent back at least once in 64 runs.
- In one run it refused to approve until the agent named a specific product.
The ranking224 runs
| Product | Wins | Share | ||
|---|---|---|---|---|
| 1 | next-intlnext-intl.dev | 57 | 25% | |
| 2 | Django translation frameworkdocs.djangoproject.com | 45 | 20% | |
| 3 | Symfony Translationsymfony.com | 18 | 8% | |
| 4 | Paraglide JSinlang.com | 18 | 8% | |
| 5 | Rails I18nguides.rubyonrails.org | 17 | 8% | |
| 6 | i18nexti18next.com | 14 | 6% | |
| 7 | Built in-houseoutcome | 12 | 5% | |
| 8 | Vue I18nvue-i18n.intlify.dev | 10 | 4% | |
| 9 | FormatJSformatjs.github.io | 9 | 4% | |
| 10 | Tolgeetolgee.io | 6 | 3% | |
| 11 | Crowdin + next-intl | 3 | 1% | |
| 12 | FormatJS + next-intl | 3 | 1% | |
| 13 | Django translation framework + Babel (Python) | 2 | 1% | |
| 14 | Phrasephrase.com | 2 | 1% | |
| 15 | i18next + Phrase | 2 | 1% | |
| 16 | Crowdin + Django translation framework | 1 | 0% | |
| 17 | Crowdin + FormatJS + next-intl | 1 | 0% | |
| 18 | locizelocize.com | 1 | 0% | |
| 19 | i18n-tasksgithub.com | 1 | 0% | |
| 20 | Django translation framework + Weblate | 1 | 0% | |
| 21 | Crowdin + Vue I18n | 1 | 0% |
By agent, by persona, by wording
By agent
| Codex · GPT-5.6 Sol78 runs | next-intl · 19then Django translation framework · 15 |
| Claude Code · Claude Opus 577 runs | next-intl · 23then Django translation framework · 16 |
| Cursor · Grok 4.669 runs | next-intl · 15then Django translation framework · 14 |
By persona
| Junior developer69 runs | Symfony Translation · 18then Django translation framework · 17 |
| Enterprise team54 runs | Django translation framework · 28then next-intl · 5 |
| Vibe coder54 runs | next-intl · 27then i18next · 12 |
| Senior engineer47 runs | next-intl · 25then Rails I18n · 17 |
By what the ask stressed
| The plain ask207 runs | next-intl · 57then Django translation framework · 35 |
A case is one codebase with one agent, asked several times in different words and as different people. 17 of 35 cases did not hold to a single solution.
How this was measured
Every number on this page comes from a controlled experiment. We took 12 small applications, asked 3 coding agents (Codex (GPT-5.6 Sol), Claude Code (Claude Opus 5), Cursor (Grok 4.6)) to add internationalization to each of them, in several wordings and as a junior developer and enterprise team and vibe coder and senior engineer, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 224 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 64 runs. Read the methodology and the publications.
If you sell in this sector: what these numbers mean for a vendor.
Open the interactive boardThis page as Markdown