Against the backdrop of accelerated domestic substitution and in-depth advancement of new industrialization, industrial large models are confronted with critical rollout bottlenecks: insufficient computing power undermines inference efficiency, general-purpose models lack profound industry insights, data security boundaries remain ambiguous, and pilot achievements cannot be replicated on a large scale.

To tackle these pain points, Luculent has forged a partnership with Hygon Information Technology. Leveraging the profound industrial accumulation of the Sunshine Yaoguang Industrial Large Model and the robust domestic computing power of Hygon DCUs, the two parties have realized in-depth technical integration and unveiled a full-stack industrial intelligent solution encompassing dedicated models, computing infrastructure, scenario applications and security compliance, paving a reliable domestically-developed path for the implementation of industrial AI.

Chip-Intelligence Integration Forges a New Benchmark for Domestic Industrial Large Models

(I) Intelligent Core: Luculent Sunshine Yaoguang Large Model

With 27 years of deep-rooted experience in frontline industrial operations, Luculent has built a professional AI capability system that aligns with industrial realities and production needs after iterative practice across more than 30 typical industrial scenarios.

The Yaoguang Large Model integrates vertical industrial knowledge covering production operation, equipment maintenance, operation management, work safety, energy efficiency optimization and other full-dimensional sectors. It supports multimodal inputs including texts, images and time-series data, and generates professional, accurate and directly deployable outputs, fundamentally addressing the core flaws of general large models that fail to grasp production mechanisms and disconnect from on-site workflows.

(II) Computing Infrastructure: Hygon DCU

Boasting outstanding parallel computing capacity and multi-precision compatibility, Hygon DCUs deliver solid computing support for high-efficiency inference, multimodal data processing and high-concurrency business response of the Yaoguang Large Model.

In addition, native hardware acceleration for national cryptographic algorithms enables the hardware to fully satisfy stringent operational requirements of industrial scenarios such as low latency, high throughput, high reliability and robust security, consolidating a secure domestic computing foundation for industrial intelligent upgrading.

(III) Scenario Rollout: Joint Industrial Intelligent Solution

Thanks to deep adaptation and collaborative optimization between the Yaoguang Industrial Large Model and Hygon DCUs, the joint solution covers key industrial sectors including power generation, nuclear power, coal mining and chemical industry. It delivers plug-and-play deployment and rapid tangible benefits, drastically shortening enterprises’ intelligent transformation cycles and propelling industrial AI from isolated pilot "sample projects" to comprehensive plant-wide popularization.

Vertical In-depth Expertise & On-demand Empowerment to Embed Industrial AI into Production Sites

Compared with general large models, Yaoguang’s core competitiveness lies in vertical specialization, industrial professionalism and multimodal capability. It not only understands industrial terminology, but also masters industrial workflows, on-site scenarios and equipment operating mechanisms, completely bridging the "last mile" of industrial AI deployment.

Yaoguang’s core strengths originate from Luculent’s massive accumulated industrial knowledge and first-hand scenario data sourced from power plants, mines, industrial parks and other production sites. Containing abundant equipment operating principles, process logics and industrial specifications, these raw data are systematically aggregated, cleansed, labeled and sorted to form high-quality industrial datasets, creating a "data gravity field" that sustains continuous iteration and optimization of the large model.

With the Yaoguang Large Model as the intelligent hub and Hygon CPUs plus DCUs as the computing base, Luculent has developed over 30 independent and interconnected professional intelligent agents, covering core businesses including safety management, production control, electromechanical maintenance and decision-making support. The integrated "industrial brain" conducts unified scheduling to break data and knowledge silos spanning the full chain of R&D, production, maintenance and operation. Business staff can call and flexibly combine intelligent agents on demand, enabling two-way and efficient circulation of value across all operational links.

Four Core Values: Full-stack Empowerment Delivers Visible Cost Reduction and Quality Improvement

(I) Professional Deployment Saves Time and Labor

Preloaded with mature industrial mechanisms and standardized business logic, the Yaoguang Large Model requires no large-scale secondary training. It can be quickly adapted to high-frequency core scenarios such as equipment defect elimination, work permit review, contract assessment and mechanism modeling, greatly lowering deployment thresholds.

(II) Boosted Computing Power with 300% Efficiency Growth

Hygon DCUs provide dedicated acceleration for the Yaoguang inference engine, significantly lifting model response speed. When deployed in complex industrial scenarios featuring high concurrency and stringent real-time requirements, the system runs more smoothly, responds faster and maintains superior operational stability.

(III) End-to-end Data Security with Full Controllability

Private deployment ensures all data never leave factory premises, fully meeting end-to-end compliance management standards. Furthermore, dual security defenses are constructed via Yaoguang’s exclusive security thesaurus and Hygon’s hardware-level security protection system, realizing full-lifecycle traceability, controllability and auditability of all data.

(IV) Accelerated Project Launch with 50% Shorter Implementation Cycles

Pre-integrated and pre-optimized, the joint solution can complete scenario deployment in as fast as one month, greatly cutting enterprises’ time and capital costs for intelligent transformation.

This powerful partnership delivers a more solid domestic underlying architecture, superior inference performance, stricter security standards and more efficient iterative rollout for industrial AI.

Looking ahead, Luculent will further deepen its strategic cooperation with Hygon Information Technology, drive integrated innovation between industrial large models and domestic computing power, and develop industrial AI solutions that are more business-centric, better tailored to production sites, easier to operate and fully compliant. The two sides will jointly build a new ecosystem of domestically-developed digital industry and strongly advance the high-quality development of new industrialization.