Code translation converts source code from one programming language to another while preserving functionality. Approaches: Rule-based: Syntax mapping rules, limited to similar languages. LLM-based: Models trained on parallel code understand semantics, generate target language. Transpilers: Specialized tools (TypeScript to JavaScript, CoffeeScript to JavaScript). Model capabilities: GPT-4/Claude handle many language pairs, specialized models like CodeT5 for translation. Challenges: Language paradigm differences (OOP vs functional), library mapping (standard libraries differ), idiom translation (natural code in target language), edge cases and language-specific features. Use cases: Legacy modernization (COBOL to Java), platform migration, polyglot codebases, learning new languages via comparison. Quality concerns: May produce non-idiomatic code, could miss language-specific optimizations, testing crucial. Evaluation: Functional correctness (does translated code work?), compilation success, test suite passing. Best practices: Translate incrementally, maintain comprehensive tests, review and refactor output, handle dependencies separately. Valuable for migration projects.