DeepSeek, the Chinese AI startup known for its open-source models, has unveiled a new generation of artificial intelligence systems that claim to significantly narrow the performance gap with leading proprietary models. The company's latest models, which are said to be more efficient and performant than their predecessors, represent a major step forward in the ongoing competition between open and closed-source AI systems.
Architectural Improvements Drive Performance Gains
The new models, which DeepSeek describes as having
