DeepSeek (深度求索), founded in 2023, is a Chinese technology company committed to advancing Artificial General Intelligence (AGI). They focus on developing cutting-edge AI models like DeepSeek-V3 and DeepSeek-R1, which boast improved reasoning, enhanced front-end capabilities, and reduced hallucinations, supporting features like JSON output and function calling. Beyond models, DeepSeek also innovates in core AI infrastructure, including high-performance distributed file systems (3FS), optimized parallelism algorithms (DualPipe, EPLB), and efficient GPU kernel libraries (DeepGEMM, DeepEP, FlashMLA). These foundational technologies are designed to optimize throughput, latency, and cost efficiency for large-scale AI training and inference workloads, demonstrating their commitment to both model development and the underlying systems that power AGI.