BeijingMay 19, 2026 /PRNewswire/ — Since April 2026, the full rollout of GPT-5.5, China’s daily AI token calls surpassing 140 trillion to top the global rankings, and the large-scale deployment of Agentic AI have converged as three major trends, propelling the AI industry into a critical inflection point for commercialization. iSoftStone is focusing on four key dimensions: AI Infrastructure, Computational Intelligence, Scenario Intelligence, and Terminal Intelligence. Recently, its computing power base first adapted to DeepSeek-V4, its server products secured a 1.753 billion yuan order from China Mobile, industry project implementations achieved penetrating supervision and benchmark status with China Reform Assets, and its terminal ecosystem events, in collaboration with Mechrevo, empowered developers. These four fronts are advancing in parallel, continuously unleashing full-stack synergies.
AI Infrastructure: Reidong AI Partners with Pingtan Ascend to First Launch DeepSeek-V4
In late April, the DeepSeek-V4 series models were officially released as open source, marking a milestone upgrade for the domestic large model ecosystem. iSoftStone’s Reidong AI Agent Cloud Platform responded swiftly, leveraging the Ascend computing power base at the Fujian Pingtan Cross-Strait Intelligent Computing Center to complete the deployment, adaptation, and API launch of the DeepSeek-V4 series models. This made it one of the first agent platforms in China to offer DeepSeek-V4 API services, signifying another major upgrade to Reidong AI’s “model aggregation” capability and providing an “out-of-the-box” domestic AI application foundation for industries such as manufacturing, finance, and government.
The Reidong AI platform adopts a four-layer security architecture, pre-integrates eight major model providers, and supports collaboration among 50 Agent avatars. The DeepSeek-V4 model, deployed on the Pingtan Ascend computing base, is now offering API services to enterprises through the platform, fully compatible with the OpenAI SDK format. Enterprises can achieve zero-migration-cost integration by simply modifying the model ID. This marks a critical step for iSoftStone in building a domestic AI application foundation.
Computational Intelligence: Building a Full-Stack Intelligent Computing Base, Winning a 1.753 Billion Yuan Order
On the computing hardware front, iSoftStone’s full-stack intelligent computing brand, iSoftStone Huafang, continues to refine its product portfolio. At the spring new product launch on April 22, iSoftStone Huafang unveiled its new brand strategy, FunAI³, centered on three pathways: “Intelligent Computing Foundation, Ecosystem Aggregation, and Industry Deepening,” to build a complete computing system from edge inference to central training.
Key products include:
- The Huafang SuperStrong A800I A3 super node product, offering 32 to 384 specifications, meeting the high-throughput, high-concurrency performance demands of industries like internet, telecom, and finance for large model inference.
- The Huafang SuperXuan 3700 AI quad-card liquid-cooled workstation, featuring dual Kunpeng 920 new-generation processors and a globally first liquid-cooled Ascend inference card thermal module, reducing NPU core temperature by over 40°C and noise by 30dB+. With a full configuration of four Atlas 300I Duo cards, it delivers up to 1120T FP16 computing power, supporting local inference of 70B large models.
- The Huafang SuperStrong A860, a new-generation domestic AI server equipped with the Atlas 350 accelerator card featuring the new Ascend 950PR processor. It offers ultra-high computing power, flexible scalability, and robust security, suitable for AI large language model training and inference, as well as AI-accelerated computing applications.
iSoftStone Huafang’s diverse intelligent computing product portfolio provides “software-hardware integrated, full-stack intelligent” solutions for industries such as government, finance, healthcare, education, and manufacturing. In terms of market validation, the company won the bid for China Mobile’s 2026-2027 PC server product centralized procurement project, with a bid amount of 1.753 billion yuan, indicating that iSoftStone Huafang’s domestic AI server products have gained large-scale recognition from a leading telecom operator.
Scenario Intelligence: Creating Benchmarks for Central Enterprise Supervision and Asset Management, Full-Stack AI Penetrates Core Business
On the scenario deployment front, iSoftStone officially launched its new-generation intelligent penetrating supervision platform, deeply integrating proprietary full-stack AI technology with ontology-driven architecture. This achieves a generational leap from “report collection + manual verification” to “real-time alerts + closed-loop management,” and has been successfully implemented at a leading central enterprise. Recently, the company secured several major projects: it won the bid for China Reform Asset Management Co., Ltd.’s AI platform construction project, undertaking the full construction of the enterprise-level AI intelligent decision-making platform for China Reform Assets. This project leverages iSoftStone’s self-developed enterprise-level agent platform, focusing on core business scenarios of central enterprise asset management, achieving a full-chain intelligent transformation from investment decisions and risk compliance to financial control and document governance, helping to create a benchmark for digital transformation in central enterprise asset management.
On the engineering side, iSoftStone, using the “AI intelligent risk control assistant” as an example, deeply deconstructs the financial-grade AI application deployment system, bridging the “last mile” from “requirements/ideas → technology → product → business value.”
Terminal Intelligence: iSoftStone Jinke × Mechrevo, imini E300 Empowers AI Native Development
On the terminal intelligence front, iSoftStone’s brand, iSoftStone Jinke, deeply collaborates with Mechrevo. Recently, the Spring Tide • Spring Shenzhen Hackathon concluded successfully, with nearly 300 young developers from universities including Tsinghua, Peking University, Hong Kong Polytechnic, and Harvard competing. iSoftStone Jinke and Mechrevo deeply empowered the event, leveraging “full-stack AI capabilities + native computing terminals + industry-level scenario openness” to create a complete loop from creativity to implementation. The jointly provided imini E300 mini AI host supports one-click deployment of local large models, native operation of AI Agents, and rapid integration with the OpenClaw engine, stably supporting multimodal inference, model training, and high-concurrency data processing—a single device serving as a “mobile AI R&D station.”
Full-Stack Synergy, Value Resonance
From Reidong AI’s first launch of DeepSeek-V4, to the Huafang SuperStrong A800I A3 and SuperXuan 3700 liquid-cooled workstation building the intelligent computing base, to winning the bid for China Reform Assets’ AI platform, releasing the penetrating supervision platform, and empowering the developer ecosystem with the imini E300—iSoftStone’s four-layer full-stack intelligent capabilities have formed synergistic effects. In 2025, iSoftStone’s AI business revenue has already exceeded half of its total, with the dividends of scale growth continuing to materialize. Currently, the pace of large-scale AI deployment continues to accelerate, and the potential of full-stack synergy will be further amplified.