At today's afternoon product technology communication session for the OPPO Find X7 series, OPPO announced its latest upgraded AndesGPT model. This model will be applied to the OPPO Find X7 series smartphones, serving as the technological backbone for the powerful smart assistant, Little B. The AndesGPT model comes in three sizes: 180 billion parameters, 70 billion parameters, and 7 billion parameters, supporting a combination of on-device and cloud deployment to meet user needs in different scenarios. The OPPO Find X7 series will be the first to feature the 7 billion parameter AndesGPT on-device model, offering users a more intelligent experience. With robust knowledge, memory, and tool capabilities, the AndesGPT model provides Little B with a broad knowledge base and personalized intelligent service capabilities. Additionally, through the innovative SwappedAttention technology, AndesGPT addresses the issue of initial inference delay caused by long-term memory, significantly improving Little B's response speed. In text content summarization scenarios, the first 200 characters can be generated in just 0.2 seconds, outperforming competitors by 20 times; the first 2,000 characters in 2.9 seconds, leading by 2.5 times; and the summary can reach up to 14,000 characters, easily accommodating a typical journal article and surpassing competitor models by 3.5 times. Compared to a 10 billion parameter model, the 7 billion parameter AndesGPT excels in comprehension. For example, in call summarization, it can help users organize key points of the conversation and accurately output a summary with themes, key points, and action items, significantly outperforming 10 billion parameter on-device models in accuracy.