Large language models (LLMs) are now leading the field of translation, outperforming traditional neural machine translation (NMT) in quality and adaptability, although they remain slower. GPT-4 is recognized as the best overall for translation due to its quality and customization capabilities, while Claude 3.5 Sonnet is noted for its strong data privacy, and Gemini 1.5 Pro excels in speed for large documents. Organizations should select translation models based on specific needs such as language coverage and customization, with options for proprietary or open-source models. LLMs can optimize translation quality through techniques like prompt injection and fine-tuning, despite challenges such as slower speeds and the risk of "hallucinations" in sensitive domains. Contentful supports multilingual content workflows by integrating translation capabilities and allowing for a mix of models, offering flexibility and avoiding vendor lock-in.