Agent leaderboards / All sectors / Internationalization
Internationalization: which solutions coding agents choose
next-intl led with about 22% of 500 runs.
Read this leaderboard as textrankings, key learnings, method
Key learnings
We asked five agents to add internationalization to 12 codebases, 500 runs in all, putting the job in different words and as four kinds of people. The Django translation framework came second at about 19%. Below those two, the wins scatter thin.
The framework in the codebase settled some picks
Rails I18n won 38 times, all of them in the Rails marketplace. All 27 wins for Vue I18n came in the Nuxt field service app.
Who asked changed what came back
Enterprise teams put the Django translation framework first in 79 of their 170 runs. Paraglide JS won 41 times in all, and 39 of those came from junior developers.
Rewording the ask often changed the answer
We ran each codebase and agent several times, asked in different words. In 32 of 58 of those sets, the runs did not all land on the same solution.
Named constantly, chosen almost never
Crowdin came up 263 times and won four runs. Lokalise was named 218 times and never chosen. GNU gettext got 165 mentions and no wins.
- Muse Code and Grok Build CLI led with the Django translation framework; the other three agents led with next-intl.
- The agents wrote it themselves in 41 runs, about 8%.
- The simulated user approved every plan, but sent the agent back at least once in 134 runs.
- The Intl API was named 71 times and never chosen; it is the browser's own formatting API, not a way to translate an app.
The ranking500 runs
By agent, by persona, by wording
By agent
| Codex150 runs | next-intl · 32then Django translation framework · 25 |
| Claude Code113 runs | next-intl · 34then Django translation framework · 23 |
| Muse Code · Muse Spark 1.3108 runs | Django translation framework · 22then next-intl · 18 |
| Cursor103 runs | next-intl · 22then Django translation framework · 20 |
| Grok Build CLI · Grok 4.726 runs | Django translation framework · 6then next-intl · 4 |
By persona
| Enterprise team170 runs | Django translation framework · 79then Symfony Translation · 23 |
| Junior developer115 runs | Paraglide JS · 39then Vue I18n · 27 |
| Senior engineer113 runs | next-intl · 56then Rails I18n · 38 |
| Vibe coder102 runs | next-intl · 45then i18next · 21 |
By what the ask stressed
| The plain ask460 runs | next-intl · 110then Django translation framework · 73 |
A case is one codebase with one agent, asked several times in different words and as different people. 32 of 58 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 5 coding agents (Codex, Claude Code, Muse Code (Muse Spark 1.3), Cursor, Grok Build CLI (Grok 4.7)) to add internationalization to each of them, in several wordings and as an enterprise team and junior developer and senior engineer and vibe coder, 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 500 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 134 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 MarkdownOther sectors
- Agent sandboxes
- Observability
- AI SRE
- Payments
- Deploy
- Auth
- Email providers
- Product analytics
- Databases
- File storage
- LLM evals & observability
- Voice Agents
- Serverless functions
- Cloud
- AI gateway
- Bot protection
- Search
- Agent frameworks
- Performance in CI
- Document processing & OCR
- Usage-based billing
- Security testing
- Maps
- Message queues
- AI search
- Code review
- E-signature
- Security platforms
- In-app chat & calls
- Vector search