DeepSeek has launched its open-source reasoning models, DeepSeek-R1 and R1-Zero, now accessible on Hugging Face under an MIT license. DeepSeek-R1, with 671 billion parameters, matches OpenAI's performance in several benchmarks, particularly in math and coding tasks. This model excels in problem-solving while ensuring its outputs are fact-checked, albeit at a potential cost to response speed. However, content restrictions from Chinese authorities apply. Meanwhile, the U.S. is considering new export restrictions on advanced AI technologies.