Guandata ChatBI Completes DeepSeek-R1 Large Model Integration and Upgrade, Embarking on a Leap in Frontier Exploration | Linear Capital Portfolio Company

|

In the large language model (LLM) application space, Grandata (观远数据) has consistently pursued an open ecosystem to drive the democratization of intelligent data analytics, demonstrating broad compatibility.

Beyond supporting mainstream international LLMs such as GPT, Grandata also enables deep integration with leading domestic model providers. Notably, since the release of DeepSeek V2.0, Grandata was among the first to pursue integration, successfully deploying it across multiple client projects.

In early February 2025, Grandata ChatBI achieved a further technical upgrade, officially completing adaptation for DeepSeek's latest R1 model.

We believe the release of DeepSeek-R1 introduces significant variables and opportunities for the deep fusion of "large models + data analytics":

- Leap in complex reasoning capabilities, reshaping analytical depth: The DeepSeek-R1 model optimizes multi-tier reasoning architecture through reinforcement learning, demonstrating powerful reasoning and analytical capabilities in complex scenarios such as analytical inference, anomaly attribution, and intelligent data interpretation.

- An economical and efficient intelligence engine, lowering barriers to entry: DeepSeek-R1's API costs are just one-fifth of GPT-4o's, substantially reducing invocation expenses. With equivalent compute, it delivers more comprehensive analytical results, enabling enterprises to achieve precise data analytics at lower computational cost.

- A new paradigm for knowledge base scenarios, accelerating business integration: The DeepSeek model comes with built-in general industry knowledge graphs, providing out-of-the-box business recognition capabilities for verticals including retail, finance, and manufacturing. When enterprises connect CRM, ERP, and other systems, the DeepSeek-R1 model automatically aligns industry-specific terminology, compressing knowledge base cold-start cycles from two weeks to three days.

Leveraging DeepSeek-R1's exceptional characteristics in complex reasoning, cost control, and knowledge base integration, Grandata ChatBI has embarked on a new chapter of intelligence. Grandata ChatBI, an intelligent data Q&A product built on large language models (LLMs), will deeply integrate with the DeepSeek-R1 model to bring transformative change to enterprise data-driven decision-making.

Grandata ChatBI is an intelligent data Q&A product built on large language models (LLMs), providing capabilities including intent recognition, knowledge retrieval, query understanding, data querying, and visualization generation. Users can obtain data analytics results through natural language queries, enabling agile decision-making. It effectively addresses the flexibility and accessibility limitations of traditional BI. The DeepSeek-R1 model enhances reasoning capabilities through reinforcement learning, supports complex task processing, and features low cost and high efficiency.

This integration of Grandata ChatBI with the DeepSeek large model, with DeepSeek-R1 as the core foundation, enables the reshaping of enterprise data decision-making with "low barriers, high precision, and full scenarios." Particularly in complex queries and cross-table analysis, it significantly improves ChatBI's response speed and accuracy.

|

- More accurate natural language-to-data queries: After introducing large models, data analytics products significantly enhance parsing of user fuzzy semantics (e.g., "analyze reasons for sales anomalies in South China"), automatically decomposing them into multi-layer analytical tasks such as metric attribution, trend comparison, and correlation factor mining, outputting structured conclusions. During conversational analysis with ChatBI, the large model can process user input in real-time, rapidly generate analytical results, and provide feedback. Simultaneously, based on user feedback—such as follow-up questions or adjusted analytical dimensions—the large model can continuously optimize analytical pathways, iteratively producing results better aligned with user expectations.

- More agile data analytics efficiency: Large models can be combined with ChatBI's localized deployment capabilities, enabling enterprises to achieve minute-level complex query responses at lower computational cost while ensuring data security.

- More personalized enterprise-grade knowledge bases: Different industries have unique business terminology and knowledge systems. Grandata ChatBI integrates industry knowledge bases—covering terminology in retail, manufacturing, finance, and other sectors—precisely identifying professional expressions in business scenarios, such as "prospect first-order conversion rate," and automatically linking to enterprise private data assets. It not only understands general business language but also precisely grasps industry-specific terminology and business logic, delivering analytics results tailored to actual business needs. Moreover, enterprise historical analytical logic (such as marketing campaign review templates) and data metric rules can be accumulated in the ChatBI knowledge base, with the large model's continuous learning mechanism ensuring automatic optimization as the business evolves, addressing the pain point where "repetitive demands account for 70%," and unlocking data team productivity.

- More secure enterprise data ecosystems: Grandata ChatBI supports hybrid cloud, private cloud, and on-premise deployment models, with sensitive data circulating only within the enterprise intranet. Government and enterprise clients prioritizing data security can leverage large model and ChatBI infrastructure capabilities for private deployment, building scenario-based conversational BI suited to intranet environments.

From being among the first to integrate the DeepSeek-V2 large model and successfully deploying it across multiple client projects, to formally completing adaptation for the DeepSeek-R1 large model in this upgrade, Grandata has continuously explored and practiced, steadily growing confidence in the deployability of large models. The deep integration of Grandata ChatBI with the DeepSeek large model further strengthens Grandata's AI+BI intelligent capabilities, delivering a more intelligent and efficient data analytics experience for users.

Meanwhile, Grandata possesses powerful compatibility and openness in large model integration. Currently, Grandata has connected and packaged various mainstream large models, covering public cloud providers, private models, and more, while also supporting custom model integration, fully satisfying diverse and personalized customer needs and providing more flexible and comprehensive options.

In 2025, Grandata will further focus on "large model + data analytics" application scenarios, continuing to invest in deep exploration and innovation to unlock the enormous potential in this domain. Grandata looks forward to collaborating with more customers and partners in co-creation projects, exchanging insights, integrating resources, and collectively unlocking the infinite possibilities hidden in this field, empowering intelligent digital operations and upgrades.

|

About Linear Capital

Linear Capital is an early-stage investment firm focused on the "frontier technology + industry" direction—that is, frontier technologies represented by data intelligence, digital new infrastructure, next-generation robotics, and technological transformation in traditional domains (such as biomedicine, materials, and energy), applied across vertical industries to substantially improve industrial efficiency, empower solutions to pain points, and complete industrial upgrading. Through significant enhancement of industrial value, we achieve excess returns in commercial value. We currently manage ten funds with total assets under management of approximately $2 billion.

Our investment stage focuses primarily on leading angel to Series A rounds, with individual investments ranging from $1 million to $10 million (or RMB equivalent).

To date, we have made early-stage investments in over 120 entrepreneurial teams including Horizon Robotics, Kujiale, Sensors Data, Tezign, Rokid, Grandata, and Agile Robots. The combined valuation of Linear's portfolio companies is approximately $20 billion. In the near term, Linear Capital is working to become the premier "Data Intelligence Technology Fund," and in the long term, gradually build itself into the most influential "Frontier Technology Application Fund."