Li Xiang Tongxue’s MindGPT 3.0: A Leap in Deep Thinking and Intelligent Interaction

Li Auto Unveils MindGPT 3.0 for Li Xiang Tongxue

Li Auto has recently rolled out a major enhancement to its intelligentassistant, Li Xiang Tongxue, with the full debut of the MindGPT 3.0 model.This move marks a significant stride in Li Auto’s AI capabilities, presentingusers with a more intelligent and efficient interaction experience.

Performance Surge in MindGPT 3.0

The MindGPT 3.0 showcases a remarkable performance boost, especially in itsdeep thINKing abilities. It can compete with industry heavyweights likeDeepSeek-V3-0324 (for short reasoning chains) and DeepSeek-R1 (for longreasoning chains), highlighting Li Auto’s substantial investment in AItechnology. Users can now access this new model for free through the Li XiangTongxue mobile app and web version, diving into a new realm of intelligentinteractions.

New Interaction Paradigm

MindGPT 3.0 doesn’t just stop at performance improvement; it brings in a novelinteraction approach. The model supports structured reasoning chain displays,enabling users to visually track Li Xiang Tongxue’s thought process, making itseem like a ‘transparent’ intelligent assistant. It also comes with reflectionand re-seARCh capabilities. After grasPing user needs, it refines answers tooffer more precise and comprehensive information. Additionally, its voiceinput comprehension and error tolerance have been greatly enhanced, accuratelyunderstanding user intent EVEn with unclear voice commands.

Handling Complex Instructions

MindGPT 3.0 shines in dealing with complex instructions. Its deep thinkingcapabilities allow for more rational task planning and better execution ofcomplex question-and-answer requests. The upgraded tool ecosystem furtherenhances user convenience. For example, real-time stock and ticketinginformation retrieval is available, enabling users to enjoy intelligentinteractions while effortlessly obtaining necessary information. Thesignificantly improved continuous conversation ability ensures anunprecedentedly smooth and natural interaction with Li Xiang Tongxue.

Irrelevant Historical Conversation Filtering

Notably, MindGPT 3.0 includes an irrelevant historical conversation filteringfeature. Li Auto’s experiments found that some open-source deep thinkingmodels, during multi-turn conversations, forcefully associate unrelated userinputs, leading to responses with irrelevant information and affecting theuser experience. MindGPT 3.0 independently analyzes historical conversations,effectively filtering out information unrelated to the current question. Thisensures response accuracy and addresses a major user concern, representing asignificant advancement for Li Xiang Tongxue in intelligent interaction.