In the Era of Agentic AI, Mininglamp Technology (2718.HK) Kicks Off Multiple Strategic Initiatives in June for Market Positioning

In June 2026, Jensen Huang's announcement of the arrival of the Agentic AI era propelled industry attention toward Agents to a new peak.

Hailed as the "First Agentic AI Stock in Hong Kong Equities," Mininglamp Technology-W (2718.HK) has attracted significant market attention. In June, the company executed a series of intensive business actions, achieving another systematic leap from technology to commerce, and from product to ecosystem.

I. On-Device Model Iteration: From a "Single Model" to a "Multi-Model Matrix"

On June 25, Mininglamp Technology officially released Mano-CUA-2.0, an on-device GUI VLA model tailor-made for AI PCs, and simultaneously launched the Thinking mode. The newly upgraded Mano-CUA-4B-2.0 is now available for download and deployment on Hugging Face and ModelScope. The company has completed two major iterations of its on-device models within just two months.

The core upgrade of Mano-CUA-2.0 shifts AI from "acting immediately upon glancing at the screen" to "thinking through multi-step procedures before taking action." In 100 testing tasks on real macOS GUI devices, the Thinking mode improved the overall task success rate by approximately 9% compared to the Fast mode, and by 10% to 13% in medium-to-high difficulty tasks. The entire process runs locally on devices, eliminating the need to upload data to the cloud.

The open-source project Mano-P has been upgraded to a dedicated series within the Mano lineup focused on on-device deployment, with more on-device models to be released continuously in the future. With this move, Mininglamp Technology has upgraded its technical system from a "single model" to a "multi-model matrix" for on-device AI, while continuing to complete the closed-loop validation of value from "technical narrative" to "commercial implementation."

IDC projects that by 2028, over 70% of enterprises globally will adopt hybrid or on-premise AI deployment architectures. In sectors such as finance, government affairs, and healthcare, data security and compliance are becoming critical factors that determine whether AI can be truly implemented. The release of Mano-CUA-2.0 hits the exact node of this structural trend.

II. Open-Source Agent Collaboration Network: Positioning for the Underlying Protocol of the "Internet of Agents"

On June 29, Mininglamp Technology officially open-sourced and released Octo, which the company defines as the "world's first open-source and trusted Agent collaboration network." Based on the Apache 2.0 license, it supports private deployment. Octo has already been deployed and implemented internally among approximately 1,400 employees, with over 2,900 AI Agents running efficiently within actual workflows.

The open-sourcing of Octo marks Mininglamp Technology's leap from "providing AI capabilities" to "defining the underlying rules of AI collaboration"—it is no longer about building individual Agents, but building the infrastructure for how Agents connect, communicate, and work together. The company has also become the only AI corporation in Hong Kong equities to possess all four layers of capability: on-device models, inference engines, Agent collaboration platforms, and hardware entry points.

In the "Implementation Opinions on the Innovative Development of 'Artificial Intelligence + Information Communications'" released by the Ministry of Industry and Information Technology in June 2026, "AI agents" were included in a national-level industrial roadmap for the first time. While leading global AI players are working on Agent collaboration, they are universally closed-source, cloud-based, and locked into their own ecosystems. Octo draws a clear line of distinction by being open-source, neutral, and privatized.

III. Entering the In-Vehicle Scenario: Partnering with PATEO Connect+ to Unlock a Trillion-RMB Incremental Market

On June 12, PATEO Connect+ (02889.HK) and Mininglamp Technology signed a strategic cooperation framework agreement. The two parties will jointly build an "AI×Car×User Demand" link platform, utilizing the Token Economy as the underlying commercial logic to bridge the complete closed loop of "brand marketing budget—in-vehicle AI computing power consumption—user scenario-based consumption." On the day the cooperation was announced, Mininglamp Technology's intraday share price surged by over 30% at one point.

This cooperation marks a critical leap for Mininglamp Technology's Agentic AI strategy from marketing, office, and content production scenarios into the trillion-RMB track of intelligent connected vehicles. It places the company on a transition path from a niche AI service provider to an incremental market participant in automotive AI agents, opening up an incremental growth space several times the size of its existing Agentic Services revenue scale.

Public data shows that the intelligent connected vehicle sector is experiencing explosive growth. The market size of intelligent driving in China reached RMB 455 billion in 2025 and is expected to further increase from RMB 580 billion to RMB 1.56 trillion between 2026 and 2030. The edge-cloud collaborative scheduling system co-built by both parties is integrating in-vehicle AI computing power consumption, Agent tasks, and brand marketing budgets into a unified value measurement and distribution system.

IV. Accelerating Global Expansion: Moving from the Hong Kong Pivot toward the European Market

On June 23, Mininglamp Technology was invited to attend the inaugural Beijing-Birmingham Technology Week, as one of only four invited Beijing-based sci-tech innovation enterprises. The event announced the "China-UK Enterprise Mutual Visit Program," granting Mininglamp Technology the opportunity to participate in the main event of Birmingham Tech Week this October. Previously, the company signed to become one of the fifth batch of key enterprise partners of the Hong Kong SAR Government and established its international headquarters at Cyberport.

Mininglamp Technology's globalization narrative is moving from its "Hong Kong pivot" toward the "European market." At Birmingham Technology Week, the company promoted its self-developed innovative achievements and expressed its willingness to expand into the UK market.

As one of the core markets of the European AI industry, the UK boasts a massive demand for enterprise-grade AI services, which is precisely the core target market for Mininglamp Technology's Agentic Services. Birmingham Tech Week is the largest regional official tech brand event in the UK, bringing together top-tier European innovative enterprises, industrial capital, and professional incubation platforms.

In addition, the company announced the launch of integration practices with AI platforms such as WeChat and Tongyi Qianwen, introducing Agent Ready solutions tailored for brand clients.

Everbright Securities (Huaxin Securities) stated that in ToB scenarios, the core value of AI applications lies in translating general model capabilities into enterprise-grade solutions to complete the "last mile." Compared to cloud vendors and pure large-model developers, the advantages that differentiate AI application companies from general model technology providers include: 1. Industry Know-How and Proprietary Data: Most of the surveyed companies previously operated traditional vertical businesses, possessing industry data accumulated over many years, deep understanding of business logic, and profound client relationships. Their familiarity with the detailed integration needs of B-end clients serves as a moat that capital cannot easily replicate quickly. 2. Customized Service Capabilities: Large domestic enterprises prefer to purchase complete solutions. Compared to the standardized platforms featured by mainstream cloud vendors, AI application service providers can arrange on-site engineers to complete complex scenario mapping, database construction, and customized system deployment, offering product forms that better align with customer demands. 3. Independence from the Ecosystems of Tech Giants, allowing for more flexible product applicability.

Currently, Mininglamp Technology's on-device model iterations, open-sourcing of its Agent collaboration network, expansion into in-vehicle scenarios, and global layout all point to a core judgment: competition in the second half of the AI industry is shifting from "building a more powerful large model" to "turning AI into organizational productivity that can execute work, collaborate, and deliver results."

As the only AI company in Hong Kong equities currently possessing all four layers of capability: on-device models, inference engines, Agent collaboration platforms, and hardware entry points, Mininglamp Technology, through its differentiated path of "Scaling Out," is evolving from the first Agentic AI stock in Hong Kong equities into a global Agentic AI infrastructure service provider.

NewTimeSpace Disclaimer: All content herein is the original work of NewTimeSpace. Any reproduction, reprinting, or use of this content in any other manner must clearly indicate the source as "NewTimeSpace". NewTimeSpace and its authorized third-party information providers strive to ensure the accuracy and reliability of the data, but do not guarantee the absolute correctness thereof. This content is for reference only and does not constitute any investment advice. All transaction risks shall be borne by the user.