Blow through the ceiling! DeepSeek epic update: APP 2.0 fully launched, countdown to V4 nuclear bomb level release! Just now, I opened the Deepseek app on my Huawei PAD and received a prompt to upgrade to version 2.0. Without hesitation, I chose to update, but the app didn't show any new features. Then I asked what new features it had, and it replied directly to me with the following content: According to the version number 2.0.0 (203) in your screenshot, this is indeed a cross generational major version update (jumping directly from 1. x to 2.0). However, the official update logs found only mentioned "fixing some known issues", which is a typical "official treasure trove of words" technique - the real changes need to be experienced by oneself. Based on industry practices and user feedback, this 2.0 upgrade is likely to include the following points: Bottom level refactoring for faster response: Jumping from version 1. x to version 2.0 usually means a complete rewrite of the underlying architecture. The intuitive feeling is that the app startup speed is faster, the delay in generating replies is reduced, and the sliding is more responsive. ️ Voice experience optimization: The "voice" option in your screenshot is still available, but considering the current trend of big models, the new version is likely to support more natural dialogue interruption or optimize the recognition accuracy of multiple languages/dialects. UI/UX Detail Fine tuning: The appearance, font, and other settings have not changed, but the internal interactive animations and visual feedback details may be optimized. · Historical bug fixes: For example, some users previously encountered issues such as unstable internet search and long conversation lag, which are likely to be resolved in this version. Although the official did not specify, this is indeed the foundational version of DeepSeek that paves the way for future big moves. Just like building a building and laying a foundation, 2.0 is designed to make the new features run more steadily in the future. ” It seems that APP 2.0 is just a prelude, the real killer move DeepSeek V4 will be released globally in late April, and every highlight will keep Nvidia up all night: -Trillion parameter MoE architecture: With a total parameter count of 1 trillion, each inference activates 37 billion parameters, fully unleashing multimodal capabilities and crushing text, images, and code in all directions! -Pure domestic computing full stack adaptation: completely bid farewell to the NVIDIA CUDA ecosystem, fully embrace the Huawei Ascend 950PR chip, train and reason independently and controllably throughout the entire chain, and break the monopoly of Western computing power! -35 times surge in inference performance: energy consumption reduced by 40%, cost only 1/70 of GPT-4, ultimate cost-effectiveness that leaves global AI giants in the dust! -Engram Memory Imprint Black Technology: Achieving "infinite context", instantly recalling the entire knowledge base, truly achieving unforgettable memories, AGI landing further! The AI industry is not boiling yet! Because the upcoming DeepSeek V4, the battle of domestic AI gods, is likely to shine brightly from the 25th to the 28th of this month, officially kicking off! The Four Major Disruptions Rewrite the Global AI Landscape, Witnessing the Historic Moment of Domestic AI Ascending to the Top of the World! Let's wait and see! Will Huang Renxun feel anxious at the DeepSeek V4 global launch event in late April? DeepSeekV4's domestically produced AI dominates the AGI revolution