Multilingual support
Overview
Claude demonstrates robust multilingual capabilities, with particularly strong performance in zero-shot tasks across languages. The model maintains consistent relative performance across both widely-spoken and lower-resource languages, making it a reliable choice for multilingual applications.
Note that Claude is capable in many languages beyond those benchmarked below. We encourage testing with any languages relevant to your specific use cases.
Performance data
Below are the zero-shot chain-of-thought evaluation scores for Claude models across different languages, shown as a percent relative to English performance (100%):
| Language | Claude Opus 4.11 | Claude Opus 41 | Claude Sonnet 4.51 | Claude Sonnet 41 | Claude Haiku 4.51 | Claude Haiku 3.5 |
|---|---|---|---|---|---|---|
| English (baseline, fixed to 100%) | 100% | 100% | 100% | 100% | 100% | 100% |
| Spanish | 98.1% | 98.0% | 98.2% | 97.5% | 96.4% | 94.6% |
| Portuguese (Brazil) | 97.8% | 97.3% | 97.8% | 97.2% | 96.1% | 94.6% |
| Italian | 97.7% | 97.5% | 97.9% | 97.3% | 96.0% | 95.0% |
| French | 97.9% | 97.7% | 97.5% | 97.1% | 95.7% | 95.3% |
| Indonesian | 97.3% | 97.2% | 97.3% | 96.2% | 94.2% | 91.2% |
| German | 97.7% | 97.1% | 97.0% | 94.7% | 94.3% | 92.5% |
| Arabic | 97.1% | 96.9% | 97.2% | 96.1% | 92.5% | 84.7% |
| Chinese (Simplified) | 97.1% | 96.7% | 96.9% | 95.9% | 94.2% | 90.9% |
| Korean | 96.6% | 96.4% | 96.7% | 95.9% | 93.3% | 89.1% |
| Japanese | 96.9% | 96.2% | 96.8% | 95.6% | 93.5% | 90.8% |
| Hindi | 96.8% | 96.7% | 96.7% | 95.8% | 92.4% | 80.1% |
| Bengali | 95.7% | 95.2% | 95.4% | 94.4% | 90.4% | 72.9% |
| Swahili | 89.8% | 89.5% | 91.1% | 87.1% | 78.3% | 64.7% |
| Yoruba | 80.3% | 78.9% | 79.7% | 76.4% | 52.7% | 46.1% |
1 With extended thinking.
These metrics are based on MMLU (Massive Multitask Language Understanding) English test sets that were translated into 14 additional languages by professional human translators, as documented in OpenAI's simple-evals repository. The use of human translators for this evaluation ensures high-quality translations, particularly important for languages with fewer digital resources.
Best practices
When working with multilingual content:
- Provide clear language context: While Claude can detect the target language automatically, explicitly stating the desired input/output language improves reliability. For enhanced fluency, you can prompt Claude to use "idiomatic speech as if it were a native speaker."
- Use native scripts: Submit text in its native script rather than transliteration for optimal results
- Consider cultural context: Effective communication often requires cultural and regional awareness beyond pure translation
We also suggest following our general prompt engineering guidelines to better improve Claude's performance.
Language support considerations
- Claude processes input and generates output in most world languages that use standard Unicode characters
- Performance varies by language, with particularly strong capabilities in widely-spoken languages
- Even in languages with fewer digital resources, Claude maintains meaningful capabilities