DeepSeek-V4 Challenges AI Giants with Open-Source Breakthrough
DeepSeek-V4 Challenges AI Giants with Open-Source Breakthrough
Chinese AI Developer DeepSeek Unveils New Language Model; Open-Source Preview of DeepSeek-V4 Released
DeepSeek-V4 Challenges AI Giants with Open-Source Breakthrough
Chinese artificial intelligence developer DeepSeek has launched its latest language model, DeepSeek-V4, the company announced on X (formerly Twitter).
The model comes in two versions: the flagship DeepSeek-V4-Pro and the more cost-efficient V4-Flash, both released under open-source licenses, according to the company. Developers claim the Pro version significantly outperforms other open-source models in world knowledge benchmarks, trailing only Google's proprietary Gemini-Pro-3.1, Reuters reports.
The release of V4 coincides with escalating tech tensions between the U.S. and China—coming just one day after the White House accused Beijing of systematically stealing intellectual property from American AI labs. Reuters notes that similar allegations were previously leveled by U.S.-based OpenAI, which claimed on April 13 that DeepSeek had used leading American AI models to train its R1 chatbot via a process called distillation, where one model learns from another's outputs. DeepSeek has denied using synthetic data generated by OpenAI.