China’s DeepSeek Releases DeepSeekV2, a 21BParameter OpenSource AI Model
DeepSeekV2, a new opensource model developed by the Chinese AI company DeepSeek, is referred to be a "intermediate step" towards nextgeneration AI sys
DeepSeekV2, a new opensource model developed by the Chinese AI company DeepSeek, is referred to be a "intermediate step" towards nextgeneration AI systems. DeepSeekV2, which has 21 billion parameters and was trained on 2 trillion tokens, strives for efficiency and performance while providing robust reasoning, coding, and multilingual task skills. According to the business, it is lightweight enough for wider deployment while competing with highend versions like GPT4 and Claude 2. In order to maximise performance and cost, DeepSeekV2 employs a Mixture of Experts design, activating only 25% of its characteristics each assignment. The announcement shows DeepSeeks intention to compete worldwide in large language model research and industry adoption, and it also highlights Chinas growing ambition in foundational AI development.
Read More..
B2B Tech News | 5 months ago