智通财经网·2026-02-04 06:26Core Insights - The report from Guolian Minsheng Securities highlights the evolution of large models from "chat tools" to "autonomous employees" in the agent era, emphasizing the importance of model unit costs as tasks become more complex and require multiple stages of interaction [1][2]. Group 1: Model Cost and Efficiency - In traditional dialogue paradigms, a single interaction requires only a few model calls, whereas workflow paradigms involve multiple stages, leading to a significant increase in model call frequency and complexity [2]. - The agent services designed for complex tasks may consume tens of times more tokens compared to basic chat, making the unit cost of models critical for scalability [2][3]. - MiniMax's M2.1 model is noted for its efficiency and cost advantages, being priced at approximately 8% of Claude Sonnet's costs, which addresses the high token cost pain points faced by developers [3]. Group 2: Long Text and Reasoning Capabilities - The M2.1 model's strong long-text capabilities allow it to handle extensive workflows, accommodating more intermediate results and reducing logical breaks due to truncation [3]. - The model is designed for automated execution and error correction, making it suitable for production systems where it can write, modify, and validate code effectively [3]. Group 3: Multi-Modal and Visual Execution - The entry of agents into office and production scenarios has shifted input sources from pure text to include visual information such as screenshots, PDFs, and tables [4]. - MiniMax's multi-modal capabilities enhance the agent's ability to understand interfaces, extract key information, and output executable steps or code, facilitating "visual-driven automation" [4]. - This capability allows for tasks such as automatic form filling, error identification from screenshots, and data extraction from charts, improving deliverability and reducing manual intervention [4].