An enterprise is building a multilingual document translation service using Claude to translate 500-word business documents between 15 language pairs. Which model capability characteristic is most important for this architectural decision?

For a multilingual translation service, the most critical architectural factor is the model's multilingual training coverage — Claude performs significantly better on high-resource language pairs (English-Spanish, English-French) than on low-resource pairs (English-Swahili, English-Yoruba). An architect must evaluate translation quality per language pair against the required 15 pairs, as capability varies substantially. Option B is incorrect — 500-word documents are approximately 700 tokens, well within any Claude model's context; window size is irrelevant. Option C (vision/OCR) is a valid preprocessing concern for scanned PDFs but is a pipeline component, not the primary capability factor for the translation system itself. Option D (tool use for glossaries) is an enhancement, not the defining capability factor.