Recently, the China Academy of Information and Communications Technology (CAICT) released the highly anticipated "General Capability Requirements for AI Development Platforms - Part 4: Large-Model Technical Requirements" (hereinafter referred to as the "Large-Model Technical Requirements"). Leveraging its technological innovations and practical application experience in the field of large-models, DataCanvas participated throughout the entire process as a core contributing unit in the development of these standards.

With the continuous accumulation of large-model technologies and experiences, the field has gradually shifted focus from technical breakthroughs to the full-process engineering implementation of development, deployment, and application. Currently, large-model development platforms are still in the early stages of development, presenting new challenges amid complex algorithms and engineering steps. However, there is a strong demand for industrial applications across various sectors. To guide orderly development in the industry, regulate self-discipline, and provide users with selection references, CAICT, in collaboration with DataCanvas, Baidu, Tencent, China Mobile, China Unicom Digital Tech, Alibaba Cloud, AsiaInfo, Ant Group, and other leading companies in the large-model field, jointly developed the Large-Model Technical Requirements.

The Large-Model Technical Requirements focus on the functionalities of large-models in AI development platforms, integrating end-to-end development and support tools, including critical components such as data processing, model development and fine-tuning, deployment and inference, and support and services. These core elements play a crucial role within the overall architecture, providing comprehensive services throughout the development of AI applications. The release of this standard offers guidance for constructing, selecting, evaluating, and accepting large-model technical capabilities in AI development platforms, aiding developers and selectors in AI platform product evaluation and reference.

DataCanvas, with its profound technical accumulation and extensive practical experience in artificial intelligence, actively drives the construction of standard systems in the AI field, fulfilling a leading role in the industry. The company has repeatedly served as a core contributing unit in formulating several AI standards, including the world's first AI model development management standard, the first commercial AI development platform, and the country's first industry-standard for large models, thus promoting the industrialization process of AI.

As a leader in AI foundational software, DataCanvas has leveraged its deep AI technical expertise to independently develop the "General + Industry" open-box large-model, DataCanvas Alaya. This model supports the processing of multimodal data and offers a series of models with different configurations and parameters, reducing the cost and barriers for large-model training and application, and catering to the needs of various developers. The large-model application outcomes developed based on the Alaya large-model—TableAgent for data analysis and Jarvex for knowledge intelligence—have been widely applied in leading business scenarios across financial, telecommunications, manufacturing, energy, real estate, and internet industries. These implementations have also been successfully included in CAICT’s first authoritative research publication focusing on large-model application, the "2023 Large-Model Application Case Collection."

Moving forward, DataCanvas will continue to leverage its leading AI technology strengths to advance the development and implementation of key standards in the AI field, contributing to the industry's high-standard, high-quality development. This will further support the application and expansion of AI in diverse sectors, enabling AI to thrive across numerous industries.