SenseTime Integrates DeepSeek Enterprise into SenseCore and Launches the Open-Source LazyLLM Framework

SenseTime announced two enhancements to its SenseCore AI infrastructure at the Global Developer Conference in Shanghai, which include the integration of DeepSeek Enterprise and the launch of the open-source LazyLLM AI development framework.

SenseCore Integrates DeepSeek Enterprise and Expands Free Edition

In February 2025, SenseCore’s WanXiang platform introduced models like DeepSeek-R1 and DeepSeek-V3, providing a free edition with 10 million tokens over three months, which was well-received by clients. To better meet growing demand, SenseTime has integrated the enterprise edition of DeepSeek, alongside localized optimizations and an expansion of computing capacity for the free edition.

To address the rising demand for DeepSeek across industries and to enhance the AI development experience, SenseCore now includes DeepSeek-R1-Enterprise, offering high performance and stability with industry-leading inference efficiency. The platform also features a comprehensive alert system to ensure reliable and efficient AI model deployment.

Additionally, SenseCore provides end-to-end enterprise AI solutions, including:

l   Custom model development: Provides lightweight model customization, leveraging reinforcement learning to enhance base models for specific application scenarios while supporting efficient inference in small models. 

l   Comprehensive AI solutions: Covers the three core modules of model customization, model development, and agent application platforms to provide seamless AI integration. 

l   Rapid full-cycle deployment: A dedicated expert team delivers end-to-end AI deployment, from business diagnostics to implementation, with the fastest deployment completed in just 14 days. 

SenseCore continues to provide enterprises and developers with highly efficient, cost-effective, and scalable AI solutions. Supporting a wide range of large models, including the SenseNova Unified Large Model, DeepSeek, Qwen, and Llama, SenseCore remains committed to delivering cutting-edge AI cloud services that enhance business intelligence and operational efficiency.

LazyLLM: A Developer-Friendly Open-Source AI Development Framework 

The growing adoption of large model agent AI applications has accelerated the demand for accessible and efficient development tools. Therefore, SenseTime launched LazyLLM, an open-source, low-code AI development framework that enables developers to build complex, customized multi-agent applications with as few as 10 lines of code. 

The key advantages of LazyLLM include:

l   Facilitates efficient iterations: SenseTime has open-sourced its accumulated expertise in AI-native development, refined through its work in large-model research and AI application development. These tools empower developers by providing a structured framework that enables rapid prototyping and deployment of high-performance multi-agent AI applications with minimal code. By offering a fast and efficient approach with low barriers to entry, LazyLLM significantly lowers the threshold for AI application development. 

l   Improves data performance: LazyLLM adopts a data flow-centric approach, enabling developers to combine existing components or integrate other open-source software through Pipeline, Parallel, and other data flow configurations. 

l   One-click deployment of complex applications: LazyLLM allows developers to seamlessly integrate AI capabilities such as intent recognition, knowledge retrieval, and large-model inference to accelerate AI-powered product development. Thereafter, the developed application can be seamlessly deployed across platforms like web applications with a single click.

l   Flexible adaptability across diverse ecosystems: LazyLLM is optimized to support various large-model services, offering a unified development experience across different foundational models, training and inference frameworks, and databases. It provides standardized invocation, deployment, and service integration, making AI development more flexible. 

For example, LazyLLM requires just 10 lines of code to implement a mathematical formula, achieving 30%-60% greater efficiency compared to frameworks like LangChain and LlamaIndex.

WanXiang: A One-Stop Large Model Platform

WanXiang is an enterprise-level large model platform based on LazyLLM that enables large-model and agent-based AI application deployment, providing an end-to-end AI solution for seamless implementation.

By leveraging a low-code development approach, WanXiang significantly reduces technical barriers to AI application development. Users without coding expertise can build private large-model applications without writing a single line of code. By structuring business logic through an intuitive drag-and-drop interface, users can leverage WanXiang to automatically generate and deploy AI applications.

WanXiang is a comprehensive model development and management platform that provides tools for model training, fine-tuning, inference and evaluation. The platform provides access to hundreds of open-source large models like SenseNova and DeepSeek, enabling businesses to rapidly iterate and deploy AI solutions. With ready-to-use AI pipelines, users are able to develop a fully operational AI application workflow in just one minute, accelerating deployment cycles.

To meet the diverse needs of different application scenarios, SenseCore provides efficient inference and fine-tuning across multiple AI models. By optimizing computational resources, the platform enhances AI performance while maintaining cost efficiency, allowing businesses to maximize AI-driven outcomes within existing infrastructure constraints. In a recent third-party API evaluation of DeepSeek-R1 published by SuperClue, SenseCore’s free edition ranked first, showcasing the platform’s robust technical capabilities.

Currently, SenseCore, through LazyLLM and WanXiang, provides comprehensive AI deployment solutions across 10 key industries, including finance, construction, and aerospace. More than 30 enterprises have leveraged these technologies to enhance AI implementation, transforming AI capabilities into real-world business productivity and digital intelligence adoption.

For more information on SenseCore: 

https://console.sensecore.cn/aistudio

For more information on LazyLLM:

Project URL: https://github.com/LazyAGI/LazyLLM

Project documentation: docs.lazyllm.ai