Benchmark reports from AI research organizations such as MLPerf

DeepSeek-R1: Chinese AI Model With 671 Billion Parameters Challenges OpenAI, 27x Cheaper

Jan 24, 2025, 06:10 AM

Chinese AI startup DeepSeek has released its latest open-source AI model, DeepSeek-R1, which is challenging industry leaders like OpenAI by offering comparable performance at a significantly lower cost. DeepSeek-R1, a reasoning model with 671 billion parameters, has been reported to outperform OpenAI's o1 in several benchmarks, including mathematics, coding, and reasoning tasks. The model was developed using less powerful chips and at a fraction of the cost typically associated with such technology, with a reported development cost of just $6 million. DeepSeek-R1 is available under an MIT license, enabling unrestricted commercial use and customization, and its API access starts at just $0.14 per million tokens, making it 27 times cheaper than OpenAI's o1. This development comes amidst growing concerns in Silicon Valley about the potential erosion of U.S. dominance in AI, as Chinese companies like DeepSeek demonstrate their ability to innovate effectively despite U.S. export restrictions on advanced semiconductors.