Chinese artificial intelligence startup DeepSeek has captured attention with its latest updates, quietly launching the new large language model, DeepSeek-V3-0324, on the Hugging Face platform. This model, weighing in at 641 gigabytes, comes with an MIT license that allows for commercial use. The latest version boasts better programming capabilities compared to its predecessor V3, positioning DeepSeek as a significant contender in the AI race.
DeepSeek reported that this update demonstrates significant improvements across many industry-standard benchmark tests, particularly shining in the American Invitational Mathematics Examination (AIME) benchmark. The new release enhances skills in coding and front-end development, providing users with smarter tool-use capabilities that streamline software development processes.
This new model from DeepSeek is expected to intensify competition with U.S. tech giants like OpenAI and Anthropic. As the company's capabilities come to the forefront with its innovative offerings, users are keenly monitoring these developments' architectures and applications. DeepSeek's initiative promises to enable exponential innovation in the AI realm, leveraging open-source measures.