Chinese AI lab DeepSeek has launched an open-source reasoning model, DeepSeek-R1, claiming it matches or exceeds OpenAI's o1 on several benchmarks, including AIME and MATH-500. The model, available under an MIT license on Hugging Face, features 671 billion parameters, indicating advanced problem-solving capabilities. While R1 is noted for its self-fact-checking ability, it faces regulatory limitations on specific topics due to Chinese regulations. DeepSeek's offering is priced significantly lower than OpenAI's, promoting wider access to powerful AI reasoning tools.