CITIC Securities: DeepSeek's next-generation new model is expected to continue the high cost-effective open-source model route.

The report from CITIC Securities pointed out that since 2026, domestic large-scale model manufacturers have been focusing on upgrading Agent and code capabilities, and competing to release new models. The upcoming next-generation new model DeepSeek is expected to continue the high-cost-effective open source model route, achieve stronger memory function and super-long context processing in terms of capabilities, improve code and Agent capabilities, and fill in the gaps in multimodal capabilities, bringing new investment opportunities in the directions of model original factory, AI applications, and AI infrastructure. 1. Model original factory: The next-generation DeepSeek model is expected to work hand in hand with other domestic models to drive Chinese AI acceleration towards the world, while further reducing the cost of model training and promotion, driving overall increase in global large model API call volume with cheaper tokens. 2. AI applications: The model platform can help alleviate the anxiety caused by the market's contradictory narratives of models and applications, helping thousands of AI Agents in various industries to be implemented, benefiting AI application companies with barriers to entry. 3. AI infrastructure: Cost reduction and increased usage brought about by growth will benefit AI infrastructure, with domestic AI infrastructure and domestic models moving in the same direction.