January 20, 2025:
DeepSeek Rivaling OpenAI with New AI Model - Chinese AI lab DeepSeek launched DeepSeek-R1, a reasoning model that reportedly outperforms OpenAI's o1 on benchmarks like AIME, MATH-500, and SWE-bench Verified. The R1 model, with 671 billion parameters, self-fact-checks to ensure reliability in complex domains but faces Chinese regulatory restrictions.
Despite being hardware-intensive, R1 has more accessible, distilled versions and costs 90%-95% less via API. Amid U.S.-China AI tensions, DeepSeek and other Chinese labs are rapidly advancing, prompting calls for U.S. AI development support.
