A China First! Shanghai Launches Urban Planning and Resources AI Large Model: SenseCore Makes Urban Governance “Smarter”

Recently, Shanghai achieved a significant milestone with the release of its Quantum Urban Spatial Intelligence Construction — “Yunyu Xingkong Large Model (Professional Edition)”. This represents China’s first basic large model in the urban planning and resources field.

Jointly developed by the Shanghai Municipal Bureau of Planning and Natural Resources and SenseCore, the model addresses existing gaps in the industry, and with its powerful capabilities to “answer all questions, retrieve maps, analyze statistics, recognize images, and generate reports”, it serves as an “AI partner” for professionals in urban and natural resources planning. It pushes urban governance a crucial step forward towards scientization, refinement, and intelligence.

As China’s first basic large model in the planning and resources field, “Yunyu Xingkong” features a “1+6” model system — anchored by one industry base model and supplemented by six vertical models. It combines general intelligent capabilities with in-depth insights into industry-specific needs, while establishing a multi-domain agent system tailored for various business objectives.

The model currently addresses 10 core scenarios across three major fields: planning and resources, government governance, and social co-creation. This facilitates the in-depth transformation of technological innovation into improved governance efficiency.

An “Expert Assistant” for Planning and Analysis

Unlike ordinary chatbots, the “Yunyu Xingkong Large Model (Professional Edition)” is an industry “expert” deeply integrated into the entire workflow of planning and resources management. It was trained on China’s first fully connected specialized corpus in the planning and resources field: the “Kunyu Jinglue · Corpus”. It has a parameter scale of up to 600 billion, offering a profound understanding of the laws covering territorial spatial planning and the evolution of natural resources management.

Supported by the SenseCore and SenseNova multimodal large model integrated system, the model can interpret cross-modal data, such as satellite remote sensing images, urban planning drawings, 3D real scenes, and text reports, and features an independent decision-making scheduling engine, integrating spatial-temporal intelligence, vertical models, and multi-agent systems to tackle complex scheduling tasks, transitioning from single-point intelligence to system-wide intelligence.

Five core professional capabilities make “Yunyu Xingkong” a capable assistant for professionals:

• “Answering All Questions”: Based on over 900 high-quality planning documents, the system has accumulated more than 100,000 sets of basic Q&As and over 1,000 industry expert Q&A pairs, building a professional knowledge base covering 10 major fields, including Shanghai’s overall planning, the 15-minute community living circle concept, industrial land use, and ecological composite utilization. This knowledge base is regularly updated in alignment with business developments. By June 2027, the construction of knowledge graphs for over 60 full professional fields in the planning and resources system is expected to be completed, providing accurate and reliable decision-making consulting services.

• “Map Retrieval”: The model simplifies traditional GIS operations, allowing users to quickly locate land parcels through natural language descriptions and call up corresponding business layers and attribute information, such as real-time photos, planning drawings, 2D/3D maps, historical features, and public service facilities. This significantly improves data query efficiency.

• “Statistical Analysis”: To provide in-depth mining of data, the model is connected to the mainframe system of the Municipal Bureau of Planning and Natural Resources. It can perform intelligent queries and statistically analyze vital business data, such as land transfers and construction conditions, and display the results in spatial correlation through “map-data linkage”, providing real-time, accurate, and intuitive data support for management decisions.

• “Image Recognition”: This feature enables professional drawings to “speak”. Integrating SenseNova multimodal large model technology, the model can identify elements in drawings and extract key spatial feature information, thereby efficiently assisting in vision-intensive tasks, such as drawing compliance reviews and current situation feature comparison.

• “Report Generation”: This is a powerful tool to improve work efficiency. Leveraging its professional knowledge base, chain-of-thought, and business data analysis capabilities, the model can understand business logic and assist in generating business reports that comply with technical specifications, balancing efficiency and quality.

Empowering 10 Scenarios to Open Up the Transformation Channel from Technology to Efficiency

In practical applications, the “Yunyu Xingkong Large Model (Professional Edition)” has demonstrated strong empowerment effects in 10 scenarios across three major fields:

• In the planning and resources field, it makes urban overall planning more intelligent, land supply more precise, and industrial land utilization more efficient, helping modernize territorial spatial governance.

• In the government governance field, it fosters scientific decision-making and refined governance in scenarios such as ecological composite utilization, business environment optimization, and high-quality municipal construction.

• In the social co-creation field, it provides support for shaping Shanghai-style Jiangnan characteristics, building 15-minute community living circles, protecting historical features, promoting public participation and social collaboration, and constructing a new pattern of co-construction, co-governance, and shared benefits.

Earlier, a public version of “Yunyu Xingkong” was launched online, providing the public with knowledge integration services based on open planning, a resources corpus, and search engine capability.

Toward Spatial Intelligence: A Leap from “Understanding Drawings” to “Urban Cognition”

As the core R&D support for the model, SenseCore provides end-to-end technical guarantees for “Yunyu Xingkong”: realizing cross-modal understanding and reasoning based on the SenseNova multimodal large model, completing complex task scheduling through an intelligent scheduling engine, and laying a solid foundation for professional cognition with a continuously optimized industry corpus.

Looking ahead, the model aims to evolve from 2D data processing to 3D spatial cognition, supported by SenseTime’s Spatial Intelligence Model SenseNova-SI. Through capabilities such as spatial measurement, spatial reconstruction, and spatial reasoning, it will be applied to scenarios such as negative space optimization, pedestrian road monitoring, and urban safety early warnings, providing a continuously evolving spatial intelligence engine for the “Quantum City” initiative.

The launch of the “Yunyu Xingkong Large Model (Professional Edition)” is an important milestone in Shanghai’s Quantum City construction. From understanding drawings to urban cognition, and from single-point empowerment to systematic support, this AI large model is redefining the working mode in the planning and resources field and injecting a steady stream of intelligent momentum into the high-quality development of urban areas.