MiniMax M2模型跻身全球前五，推动中国AI“普惠”升级

Core Insights - MiniMax has launched its new text model MiniMax-M2, which ranks among the top five globally and first in open-source according to the Artificial Analysis (AA) leaderboard, competing with major players like OpenAI and Google [1][2] - The M2 model offers significant cost advantages in its API pricing compared to international competitors, positioning itself as a "high intelligence + low cost" solution in the competitive landscape of large models [1] Performance Metrics - M2 excels in key tasks such as coding, instruction following, and agent performance, achieving top rankings in these areas [2] - The model is designed for end-to-end development workflows and demonstrates outstanding capabilities in various applications, including complex tool invocation and execution [5] Cost Efficiency - M2's API pricing is set at $0.3 per million tokens for input and $1.2 for output, making it approximately 8% of the cost of Claude Sonnet 4.5, while also providing nearly double the inference speed [6] - The model has received positive feedback from international developers, with significant usage on platforms like OpenRouter shortly after its launch [6] Market Positioning - MiniMax aims to democratize AI access for developers and small businesses, positioning M2 as a foundational model for the integration of AI into various industries [7] - The model is optimized for coding and agent tasks, enhancing efficiency and responsiveness in multi-agent workflows [7] Industry Applications - M2 is expected to play a crucial role in the digital transformation of various sectors, including finance, industry, healthcare, and education, by providing advanced capabilities in code generation and tool invocation [8] - Specific applications include improved report analysis in finance and optimized production processes in industrial settings [8] Accessibility Initiatives - To promote widespread adoption, MiniMax is offering free global API access for the first two weeks post-launch and has introduced the MiniMax Agent for various use cases [9] - The MiniMax Agent features two modes: an efficient mode for lightweight tasks and a professional mode for complex requirements, available for free on web and app platforms [9]