Japanese
How to get natural Japanese out of Syntic models, and what to check before you ship. The general rules for every language are on Work in other languages.
At a glance
| Language tags | ja-JP |
| Script and direction | Kanji, hiragana and katakana, left to right (vertical in print), no spaces |
| Token use | Measure with token counting on your own text; non-Latin scripts and long compounds usually take more tokens per sentence than English. |
Register and form of address
Politeness level is part of the grammar. Customer service normally uses です/ます (teineigo) with keigo for the company’s own actions; plain form reads as rude in that context. Specify the level.
Numbers, dates and currency
Full-width punctuation (、。). Dates: 2026年11月5日; the Japanese era (令和8年) appears on official forms. Currency ¥ without decimals.
Common pitfalls
- Length: count characters. Japanese replies that translate English length limits word-for-word come out too long.
- Katakana loanwords vs native words (ユーザー vs 利用者): give your glossary.
- Honorific suffixes: address customers with 様 after their name in support, never さん for a company-to-customer message.
A system prompt line that works
Writing the language instruction in Japanese itself is a strong signal:
日本語で、丁寧語（です・ます調）を使い、簡潔に回答してください。お客様の名前には「様」を付けてください。(Answer in Japanese using polite form, concisely; add 様 to the customer’s name.)
Keep the rest of your instructions in whichever language your team maintains best; the model follows mixed-language system prompts reliably, and the reply language follows this line.
Test before launch
- Collect 30 or more real Japanese messages, including short, misspelt and mixed-language ones.
- Have a fluent speaker grade replies for correctness, register and naturalness; then calibrate a rubric grader on their grades (Evaluate).
- Check rendering in your actual UI: line breaks, truncation and fonts.
- Re-run the set whenever you change the prompt or the model.