DeepSeek Open-Sources New V3.1-Base AI Model with 685 Billion Parameters
On August 20, Chinese artificial intelligence company DeepSeek announced the open-source release of its latest large language model, V3.1-Base, on the Hugging Face platform. This significant update introduces a model architecture featuring approximately 685 billion parameters, marking a substantial advancement in computational scale and capability. A key enhancement in this version is the extension of the context length to 128K tokens, allowing the model to process and retain significantly larger amounts of information within a single interaction. Prior to the public open-source announcement, DeepSeek had already upgraded its online service to the V3.1 version earlier that evening, notifying users of the improved context capabilities. This move underscores DeepSeek's continued commitment to contributing to the global open-source AI community while competing with major proprietary models from industry leaders. The release of the V3.1-Base model provides developers and researchers with access to a powerful tool for building advanced AI applications, potentially accelerating innovation in natural language processing and complex reasoning tasks. The event highlights the rapid pace of development in the AI sector, particularly among emerging players aiming to rival established tech giants.
Wire timeline
DeepSeek Open-Sources New V3.1-Base AI Model with 685 Billion Parameters
On August 20, Chinese artificial intelligence company DeepSeek announced the open-source release of its latest large language model, V3.1-Base, on the Hugging Face platform. This significant update introduces a model architecture featuring approximately 685 billion parameters, marking a substantial advancement in computational scale and capability. A key enhancement in this version is the extension of the context length to 128K tokens, allowing the model to process and retain significantly larger amounts of information within a single interaction. Prior to the public open-source announcement, DeepSeek had already upgraded its online service to the V3.1 version earlier that evening, notifying users of the improved context capabilities. This move underscores DeepSeek's continued commitment to contributing to the global open-source AI community while competing with major proprietary models from industry leaders. The release of the V3.1-Base model provides developers and researchers with access to a powerful tool for building advanced AI applications, potentially accelerating innovation in natural language processing and complex reasoning tasks. The event highlights the rapid pace of development in the AI sector, particularly among emerging players aiming to rival established tech giants.
TechNode