Search
DeepSeek has recently launched version V3.1 of its model, marking significant enhancements in reasoning prowess, context handling, and multilingual support. Boasting an impressive 560 billion parameters, this new iteration is adept at tackling intricate tasks, delivering a remarkable 43% improvement in multi-step reasoning performance. Its context processing capabilities have been fortified, enabling it to manage texts stretching up to 128k in length—akin to a full-fledged medium-sized novel, spanning between 100,000 and 130,000 Chinese characters. Furthermore, V3.1 boasts improved information accuracy, minimizing the occurrence of false information, and supports over 100 languages. It's worth noting that the highly anticipated DeepSeek R2 version has encountered training-related delays, pushing back its release date.
-
C114 Communication Network -
Communication Home
