Text Translation with Claude API — Python

How to translate text between 100+ languages using the Claude API in Python. Covers single-string translation, batch translation, language detection, and quality-controlled output with Pydantic.

Claude translates between 100+ languages with strong contextual understanding — it handles idioms, technical terminology, and register (formal/informal) better than pure machine translation systems. This guide covers every common translation pattern in Python.

Installation

pip install anthropic

Minimal translation

import anthropic

client = anthropic.Anthropic()

def translate(text: str, target_language: str) -> str:
    message = client.messages.create(
        model="claude-haiku-4-5-20251001",
        max_tokens=2048,
        system=f"You are a professional translator. Translate the user's text into {target_language}. Return only the translated text, no explanation.",
        messages=[{"role": "user", "content": text}]
    )
    return message.content[0].text

result = translate("Hello, how are you today?", "Spanish")
print(result)  # "Hola, ¿cómo estás hoy?"

Structured output: translation + detected language

import json

def translate_with_metadata(text: str, target_language: str) -> dict:
    message = client.messages.create(
        model="claude-haiku-4-5-20251001",
        max_tokens=2048,
        system=(
            "You are a professional translator. Return a JSON object with keys: "
            "'source_language' (detected), 'translation' (translated text), "
            "'confidence' ('high'/'medium'/'low'). No markdown fences."
        ),
        messages=[{"role": "user", "content": f"Translate to {target_language}:

{text}"}]
    )
    return json.loads(message.content[0].text)

result = translate_with_metadata("Bonjour le monde", "English")
# {"source_language": "French", "translation": "Hello, world", "confidence": "high"}

Preserve HTML/Markdown formatting

def translate_markup(html: str, target_language: str) -> str:
    message = client.messages.create(
        model="claude-haiku-4-5-20251001",
        max_tokens=4096,
        system=(
            f"Translate human-readable text to {target_language}. "
            "Preserve all HTML tags, attributes, href values, and Markdown syntax exactly. "
            "Return only the translated markup."
        ),
        messages=[{"role": "user", "content": html}]
    )
    return message.content[0].text

html_input = "
Welcome to our platform. Learn more.
"
print(translate_markup(html_input, "German"))

Batch translation (50% cheaper with Batch API)

strings = [
    "Save changes",
    "Cancel",
    "Are you sure you want to delete this item?",
    "Profile updated successfully.",
]
target = "Japanese"

requests = [
    {
        "custom_id": f"str-{i}",
        "params": {
            "model": "claude-haiku-4-5-20251001",
            "max_tokens": 512,
            "system": f"Translate to {target}. Return only the translation.",
            "messages": [{"role": "user", "content": s}]
        }
    }
    for i, s in enumerate(strings)
]

batch = client.messages.batches.create(requests=requests)
print(f"Batch submitted: {batch.id}")  # poll within 24h for results at 50% cost

Domain-specific translation with glossary

MEDICAL_GLOSSARY = {
    "myocardial infarction": "心肌梗塞",
    "hypertension": "高血压",
    "electrocardiogram": "心电图",
}

glossary_text = "
".join(f"{k} → {v}" for k, v in MEDICAL_GLOSSARY.items())

def translate_medical(text: str) -> str:
    message = client.messages.create(
        model="claude-sonnet-4-6",  # use Sonnet for critical medical content
        max_tokens=4096,
        system=(
            f"Translate medical text to Simplified Chinese.
"
            f"Apply this glossary exactly:
{glossary_text}
"
            "Preserve all medical term precision."
        ),
        messages=[{"role": "user", "content": text}]
    )
    return message.content[0].text

Claude vs dedicated MT APIs

For bulk translation jobs, estimate costs with the Claude API Cost Calculator. For text summarization across languages, see the summarization guide.

Frequently asked questions

Free tools

More examples