DeepSeek v4 출시

DeepSeek v4 launch presents two optimized model variants—the high-capacity V4-Pro and efficient V4-Flash—demonstrating the viability of MoE architectures in competing with proprietary frontier models. This release intensifies the open-source vs. proprietary divide in large language models, offering competitive alternatives to GPT-4o and Claude Sonnet.