Overview
ABot-Earth is a **City Generation Model v0.5** released by AutoNavi on June 8, 2026, positioned as the **world's first 3D-native city world model**. Unlike the traditional indirect path of "2D distillation to 3D," ABot-Earth adopts the **3DGS (3D Gaussian Splatting) native path**—directly generating 3DGS-format urban 3D scenes from input (satellite images or text descriptions), with efficiency **1000 times higher** than traditional methods.
Its core capability is **generating kilometer-level 3D city scenes on a single consumer-grade GPU in 10 minutes**: input a satellite image or a text description (e.g., "Shanghai Lujiazui Financial District, evening") to generate 3DGS city assets directly usable in Unity/Unreal engines. With kilometer-level **continuous generation** (seamless stitching of multiple segments), it can construct entire urban districts or even city-level worlds.
This is a key infrastructure at the intersection of four tracks: **embodied intelligence + autonomous driving + digital twin + AI gaming**. BEV/3DGS/city-level 3D assets were once the core moat of AutoNavi's business, now packaged into a **free world model for developers**—essentially, AutoNavi is releasing its years of city data assets externally through AI.
Key Features
- 3DGS Native Path: Bypasses the traditional indirect path of "2D distillation to 3D," directly generating 3DGS-format urban 3D scenes, with efficiency 1000 times higher than traditional methods
- Single Consumer GPU, 10-Minute Generation: No server cluster needed; a single consumer-grade GPU can generate kilometer-level 3D cities in 10 minutes, extremely low barrier to entry
- Satellite Image / Text Dual Input: Input can be a satellite image (AutoNavi's core data asset) or a text description, covering both creative and data-driven workflows
- Kilometer-Level Continuous Generation: Supports continuous kilometer-level 3D city generation with seamless multi-segment stitching, enabling entire urban district or city-level world construction
- Unity / Unreal Direct Output: Outputs 3DGS format directly usable in mainstream game engines like Unity/Unreal, no manual conversion needed
- Free Access: Free for developers, a landmark product where AutoNavi releases its core data assets externally in AI form
Use Cases
- AI game development: rapid construction of open-world scenes
- Autonomous driving simulation: generation of city-level road and scenario training data
- Embodied intelligence: virtual training environments for robot navigation and interaction
- Digital twin: urban planning, emergency drills, smart city visualization
- Film / advertising / VR content: city scene previz and background assets
- Architecture / real estate / cultural tourism: project display and virtual real-world experiences
Pros
- 3DGS native path is a generational leap in technical route, with 1000x efficiency improvement
- Usable on a single consumer-grade GPU in 10 minutes, extremely low barrier to entry
- AI-powered release of AutoNavi's city data assets, unique in China
- Direct output to Unity/Unreal, seamless integration with mainstream engines
- Free access, highly developer-friendly
Pricing
**ABot-Earth 0.5 is currently free** for developers to use. Apply for a trial at abot-earth.amap.com. Specific pricing for commercial scenarios (autonomous driving training data, enterprise-level digital twins, AI game batch production) is subject to official announcements from AutoNavi.
Summary
ABot-Earth is one of the most notable domestic achievements in the 2026 'world model' track—the combination of **3DGS native path + 1000x efficiency + single GPU 10 minutes + Unity/Unreal direct output + free** gives it clear irreplaceability at the intersection of AI gaming / autonomous driving simulation / digital twin / embodied intelligence. If you work on open-world games, autonomous driving simulation, city-level digital twins, or robot training environments, ABot-Earth is almost a must-try tool; if you only need 3D generation for a single building or indoor scene, consider lightweight tools like Meshy. This also marks AutoNavi's AI-powered release of its years of city data assets externally, a representative case of the 'data → model' assetization path.