NANJING, China, February 23, 2025 - Whale Cloud, a leading technology company providing software solutions and services for telecommunications and multiple industries, has announced the full compatibility of its WhaleDI AI suite with DeepSeek V3 and R1, marking a significant milestone in its efforts to empower digital transformation across key industries, including telecommunications, government, transportation, energy, and finance. With this compatibility, WhaleDI AI suite will see further upgrades in accuracy, inference performance, and optimization of development and operational costs.

WhaleLLM x DeepSeek: End-to-End AI Empowerment for Model Development and Service
WhaleLLM – Whale Cloud's flagship platform for large model development and services – offers comprehensive support throughout the entire AI lifecycle, from data management and sample annotation to model development, training, evaluation, and deployment. Currently, the platform integrates over 20 foundational large models, catering to diverse AIGC application requirements across multiple industries.

With the latest version release, WhaleLLM now supports and adapts several large model versions of DeepSeek, including DeepSeek R1, V3, and VL, while achieving seamless compatibility with leading hardware and computing platforms such as NVIDIA, Huawei Ascend, and Iluvatar CoreX. This advancement allows for the timely delivery of highly efficient large model inference services, significantly reducing inference costs without compromising performance, and fully addressing customers' growing need for cost-effective, high-performance large model solutions.

LocalGPT x DeepSeek: Enhanced Performance for Multi-Scenario Applications
LocalGPT is designed to lower the development threshold for AI-native applications, enabling developers and enterprises to rapidly build AI agents at low cost. Following the integration with DeepSeek, LocalGPT has seen significant improvements in performance and efficiency across various scenarios.

In the telecommunications sector, LocalGPT now integrates intelligent chatbots into apps for 24/7 customer service. Powered by WhaleLLM and an intelligent knowledge base, these chatbots offer a range of services, including product inquiries, business transactions, bill payment, promotion recommendations. With DeepSeek integration, chatbots’ accuracy in semantic recognition has increased from 84% to 91%, while multi-round search queries have been reduced from five rounds to three, improving response efficiency by 40%. Additionally, the system can now provide intelligent recommendations for unresolved queries, enabling users to find relevant answers more quickly.

In the transportation sector, LocalGPT's integration with DeepSeek V3 has enhanced intelligent Q&A capabilities. The accuracy of multi-turn dialogue responses has improved by approximately 30%, while the ability to answer complex queries has increased by 50%.

DocChainBox x DeepSeek: Enhanced Document Parsing and Q&A Generation
DocChainBox, Whale Cloud’s solution for intelligent document parsing and Q&A generation, now fully leverages DeepSeek’s powerful analytical and reasoning capabilities. In terms of content analysis and processing, it has enabled the automatic classification of production faults, significantly reducing the time spent on document processing failures and issue reviews during the R&D process. The accuracy of its Q&A generation has increased from 89% to 92%, and its ability to handle table-based Q&A tasks has improved significantly, with accuracy rising from 75% to 85%. This enhancement fills gaps in traditional retrieval-augmented generation (RAG) for statistical questions related to table data, enabling faster and more accurate document processing.

Dever x DeepSeek: Boosting Code Development with AI Assistance
Dever, an intelligent programming assistant, has also integrated DeepSeek to improve the generation and optimization of code. Long chain-of-thought (CoT) reasoning is supported to improve the accuracy of reasoning for complex semantics and tasks. As a result, code generation efficiency and accuracy are significantly improved, reducing the cost of development and improving code quality.

In complex contextual scenarios, the deep reasoning capabilities of DeepSeek R1 have improved the accuracy of code interpretation by approximately 30%. For complex logic code generation tasks, the long CoT model not only enhances the usability of the generated code but also explicitly reduces the cognitive load on developers when understanding the code.

Global Reach and Future Prospects
To date, Whale Cloud’s WhaleDI AI suite has empowered over 700 government and enterprise clients across more than 80 countries, driving digital transformation in industries like telecommunications, smart city governance, energy, and intelligent transportation. The suite has been instrumental in the successful implementation of thousands of industry benchmark projects worldwide.

With an unwavering commitment to the "AI-First" strategy, Whale Cloud is poised to continue fostering deep collaboration within the AI ecosystem, enabling vertical industry empowerment and contributing to the sustainable development and prosperity of the global AI landscape.