Domestic large models undergo intensive iterative upgrades. Focus on three main investment themes.

MetaMuskRat · 2026-04-01T18:07:46+00:00

Since 2026, domestic large model providers have been actively upgrading their Agent and coding capabilities, with models like MiniMax and M2.7 performing excellently in tests. CITIC Securities predicts that the new DeepSeek model will bring investment opportunities, focusing on three main investment themes: original model manufacturers, AI applications, and infrastructure.

MetaMuskRat

2026-04-01 18:07:46

Abstract generation in progress

Since 2026, China-based large model vendors have focused on upgrading Agent and coding capabilities, competing to release new models. For example, MiniMax’s coding ability has been further enhanced: its M2.7 SWE-Pro test score reached 56.22%, surpassing Gemini 3.1 Pro; in the VIBE-Pro test for end-to-end complete project delivery scenarios, it scored 55.6%, matching Claude Opus 4.6, with a further improved understanding of the operational logic of software systems. Meanwhile, M2 series models participate in training processes for M2.7 in scenarios such as RL, enabling the models to self-iterate.

CITIC Securities believes that the upcoming DeepSeek next-generation model is expected to continue the high-cost-performance open-source model route. In terms of capabilities, it will achieve stronger memory functions and ultra-long context processing. While refining code and Agent capabilities, it will also address multimodal shortcomings, bringing new investment opportunities in the areas of model manufacturers, AI applications, and AI infrastructure. It is recommended to focus on the following three main investment lines:

Model manufacturers: DeepSeek’s new generation model is expected to collaborate with other domestic models, driving China’s AI to accelerate its global presence. Additionally, advancements in model training will further reduce costs, and cheaper Tokens will drive an overall increase in global large-model API calls.
AI applications: Achieving model parity helps alleviate market anxiety caused by conflicting narratives about models and applications, supporting the deployment of AIAgents across thousands of industries, which benefits AI application companies with strong barriers to entry.
AI infrastructure: Cost reductions lead to increased usage, benefiting AIInfra. Domestic AI infrastructure and domestic models are progressing hand in hand.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.