Amap（高德地图）, under Alibaba Group, recently released the world's first AI-native map application: Amap 2025. This upgrade shifts traditional navigation toward 3D spatial understanding.
By analyzing multimodal data—visual, audio, textual, and positional—it interprets 3D structures of real-world environments and tracks behaviors of key elements (people, vehicles, roads, stores). Leveraging spatiotemporal reasoning, it predicts changes, activates diverse tools, and delivers optimal decisions or plans to users.
The core interactive module, "Xiaogao Teacher（小高老师）" uses a tiered computing architecture. It processes voice, text, and image inputs: lightweight models handle instant queries, while complex itineraries trigger deep processing via large models. Technically, it coordinates sub-agents including the "Travel Service Agent" (navigation), "Lifestyle Service Agent" (daily needs), and "Spatial Service Agent" (virtual-reality integration).
For driving assistance, the update integrates Beidou satellite positioning, road surveillance imagery, and real-time traffic data, offering three features: lane-departure warnings, traffic-light countdowns, and blind-spot collision alerts. The lifestyle section adds contextual recommendations—"AI Now（AI即刻）" suggests nearby services (e.g., restaurants at dusk), while "AI Explore（AI探索）" recommends cross-region destinations based on user history (e.g., weekend family stays).
The experimental "AR Check-in（AR打卡）" feature is live in select cities. Using phone cameras, it overlays virtual labels on physical landmarks for guided tours. To optimize computing resources, a hybrid offline/online architecture processes 80% of routine requests via mid/small-scale models. Some smart glasses and automakers have integrated its environmental perception modules.