China has launched DeepSeek-Prover-V2, a large language model featuring 671 billion parameters, marking a notable advancement in artificial intelligence. This model is part of the DeepSeek series, with the next iteration, DeepSeek R2, anticipated to be released soon. Reports suggest that DeepSeek R2 will significantly surpass current AI models, boasting 1.2 trillion parameters, making it approximately 10 times larger than OpenAI's GPT-4. Additionally, DeepSeek R2 is expected to be 97% cheaper to operate compared to GPT-4. The model has been trained on an extensive dataset estimated at 5.2 petabytes, indicating a substantial scale in data processing capabilities. Industry observers note that DeepSeek R2 aims to compete directly with existing AI technologies, including ChatGPT-4o, with expectations of enhanced performance and cost efficiency.