Recently, leveraging the Intelligent Computing Center jointly established with Huawei, our university successfully completed the local deployment of the DeepSeek-V4 Pro large model and made it available for full-scenario use by all faculty and students. This makes our university one of the first in the nation to achieve on-campus deployment of this flagship model, marking a new phase in the development of our AI education ecosystem.
Top-Tier Model Building a Powerful Computing Foundation
The DeepSeek-V4 series was officially released and open-sourced on April 24 this year, featuring two versions: the high-performance flagship V4-Pro and the cost-effective V4-Flash. The V4-Pro, as the flagship model, boasts a total of 1.6 trillion parameters and adopts a Mixture-of-Experts (MoE) architecture, activating only about 49 billion parameters per inference. This delivers top-tier performance while maintaining relatively economical inference costs. The model achieves an ultra-long context window of 1 million tokens. It has achieved comprehensive leadership in areas such as AI agent collaboration, world knowledge, and complex logical reasoning. Its evaluation results surpass all other open-source models, and its inference performance rivals top global closed-source products. The model efficiently supports high-intensity tasks for faculty and students, including complex scientific computing, ultra-long document processing, and multi-turn deep conversations.
Continuous Cultivation Yields Results in "AI + Education"
As a pioneer in digital transformation of education, our university continues to deepen its efforts in "AI + Education." We have established the "Intelligent Engine · Deep Space" Computing Center, creating a leading domestic AI education ecosystem. Previously, in April 2025, the university completed the local deployment of the DeepSeek R1 and V3 models. AI agents have already entered classrooms, creating a new three-in-one smart teaching model combining "digital human lecturing + intelligent practice + personalized tutoring." This transforms classrooms from passive listening to active exploration and moves knowledge from abstract memory to practical application.
Local Deployment Empowering Full-Scenario Smart Applications
Compared to cloud-based access, local deployment offers significant advantages, including higher data security, faster response speeds, and greater customization. Based on the "Intelligent Engine · Deep Space" Computing Center, the university has fully deployed this top-tier domestic large model with 1.6 trillion parameters on campus. This ensures that teaching and research data never leave the campus, fully protecting privacy and information security, while also supporting customized development and algorithm optimization.
Currently, the DeepSeek-V4 Pro large model has been fully integrated into the university's smart teaching platform, AI Agent Plaza, and "Shen Gong Xiao Zhi" (Shenyang Institute of Technology Smart Assistant) among other platforms (APP features will be updated soon). It is deeply serving three core scenarios: teaching, research, and management, and is the first to be made openly available to all faculty and students.
In the future, the university will continue to focus on a domestically developed and controllable AI technology road map, supported by cutting-edge AI capabilities such as DeepSeek-V4 Pro. We aim to drive the in-depth development of educational digital transformation, build an intelligent curriculum system, cross-modal innovative tool chains, and industry-education integration collaboration platforms, forming a replicable and scalable "Shenyang Institute of Technology Model." This will inject new momentum into serving the national AI development strategy and promoting high-quality regional economic and social development.
Translated by Basic Courses Department
