How a top Chinese AI model overcame US sanctions

With a new reasoning model that matches the performance of ChatGPT o1, DeepSeek managed to turn restrictions into innovation.