The release of DeepSeek’s V3 large language model (LLM) in January 2025 sparked an existential crisis for AI developers based on reports of very low development and training costs. To evaluate whether it is a game changer for translation, we put it through its paces with tests in three core tasks: translation, terminology extraction, and translation quality ranking. This report discusses the results and their implications for the language sector: How does DeepSeek V3 compare to other state-of-the-art LLMs? Is it a viable replacement for current translation processes? Where is it most likely to perform well?                                         Related Research

• Is ChatGPT-4o the ‘Killer App’ for Translation? 
• The Ethics of Generative AI 
• Is Generative AI’s Translation Output Usable and for What?