AI Translation API

Translate text between languages using Qwen3 and DeepSeek-V3 via the Runcrate API. Single-text and batch translation examples.

Translate text between languages using multilingual LLMs. Context-aware translations that outperform traditional MT on nuanced content.

Recommended models

Model	Strengths
Qwen3-Max	Best for CJK (Chinese, Japanese, Korean), strong European
DeepSeek-V3	Strong across all major languages, cost-effective for volume

---

Basic translation

Python

from runcrate import Runcrate

client = Runcrate(api_key="rc_live_YOUR_API_KEY")
response = client.models.chat_completion(
    model="Qwen/Qwen3-Max",
    messages=[
        {"role": "system", "content": "Translate to Japanese. Output only the translation."},
        {"role": "user", "content": "Our new API reduces integration time from weeks to hours."}
    ],
)
print(response.choices[0].message.content)

---

Multi-language

source = "Your account has been upgraded. You now have access to all premium features."
for lang in ["Japanese", "Spanish", "German", "Korean"]:
    response = client.models.chat_completion(
        model="Qwen/Qwen3-Max",
        messages=[
            {"role": "system", "content": f"Translate to {lang}. Output only the translation."},
            {"role": "user", "content": source}
        ],
    )
    print(f"{lang}: {response.choices[0].message.content}")

---

Batch — localization file

import json
strings = {"welcome": "Welcome back", "upgrade": "Upgrade your plan", "settings.save": "Save changes"}

def translate_batch(strings, target):
    r = client.models.chat_completion(
        model="deepseek-ai/DeepSeek-V3",
        messages=[
            {"role": "system", "content": "Translate UI strings. Keep JSON keys unchanged. Valid JSON only."},
            {"role": "user", "content": f"Translate values to {target}:\n\n{json.dumps(strings, indent=2)}"}
        ],
        response_format={"type": "json_object"},
    )
    return json.loads(r.choices[0].message.content)

for code, name in [("ja", "Japanese"), ("es", "Spanish"), ("de", "German")]:
    with open(f"locales/{code}.json", "w") as f:
        json.dump(translate_batch(strings, name), f, indent=2, ensure_ascii=False)

---

Context-aware translation

response = client.models.chat_completion(
    model="Qwen/Qwen3-Max",
    messages=[
        {"role": "system", "content": "Cloud computing translator. Translate to German. Keep API, SDK in English. Formal Sie."},
        {"role": "user", "content": "Deploy your model to a GPU instance in seconds."}
    ],
)

---

Tips

• Qwen3 for CJK. More natural Chinese, Japanese, Korean output.
• DeepSeek-V3 for volume. Cheaper per token, strong European languages.