I. Capital Market Layout and Industry Competitive Landscape The domestic Large Language Model (LLM) industry is accelerating its deep integration with the capital market, with leading artificial intelligence enterprises successively advancing their IPO processes to support long-term development. On May 29, 2026, MiniMax officially submitted IPO tutoring materials to regulatory authorities, comprehensively initiating its A-share listing application, and selected CITIC Securities as its exclusive tutoring institution. Prior to this, the company had completed its Hong Kong stock market listing in early 2026, and has now officially launched its strategic layout of a dual listing in both markets.
This move has completely ignited the competitive frenzy within the domestic AI track. Zhipu AI has similarly advanced its A-share listing procedures. The two leading enterprises are competing head-to-head, vying for the industry title of the "First A-Share LLM Stock." Successfully landing on the A-share market will bring a stable capital influx to MiniMax, comprehensively supporting the company's technological R&D, high-end talent recruitment, and domestic and overseas market expansion. Concurrently, this move can effectively enhance its domestic brand influence and deepen in-depth cooperation with local enterprises and public institutions, serving as a core strategic measure for MiniMax to consolidate its domestic market position.
II. Operational Data and Annual Financial Performance Since its listing on the Hong Kong stock exchange in January 2026, MiniMax's market valuation has achieved leapfrog growth. As of May 29, the company's stock price climbed to HKD 840, surging 409.09% from its IPO issue price, with a total market capitalization reaching HKD 263.454 billion. This stellar capital market performance relies on the steady expansion of the enterprise's globalized business and robust commercialization capabilities.
Over the past two months, the enterprise's Annual Recurring Revenue (ARR) has consistently maintained a growth rate of over 100%, fully validating that its AI products and services have gained widespread market recognition. At the client level, the total number of enterprise clients and developer users has broken through the one million mark, achieving a five-fold increase within half a year. Its business covers more than 200 countries and regions globally, with a cumulative global user base reaching 300 million. It is worth noting that over 70% of the enterprise's revenue comes from overseas markets, indicating significant practical results from its global layout strategy.
The financial data for the fiscal year 2025 further reflects the continuous optimization of the enterprise's operational efficiency. Last year, the enterprise's total revenue reached USD 79.038 million, of which self-developed AI-native product revenue was USD 53.075 million, and open platform and enterprise customized service revenue was USD 25.963 million. The overall gross margin increased to 25.4%, and the adjusted net loss narrowed to USD 250 million. All core metrics indicate that the enterprise's commercialization deployment is steadily advancing, and its profitability continues to improve.
III. Product Iteration Upgrades and AI Agent Capability Optimization In 2026, MiniMax continued to enrich its product matrix, launching three brand-new large language models during the year and open-sourcing two models to the developer community, further perfecting its product ecosystem. Among them, the MiniMax-M2.5 model has gained the favor and widespread use of developers by virtue of its ultra-high cost-effectiveness. However, global industry competition is becoming increasingly white-hot, and the brand's ranking on the mainstream international AI model trading platform, OpenRouter, has dropped out of the top ten, highlighting the fierce competitive landscape of the global LLM track.
In addition to foundational large language models, the enterprise completed a comprehensive upgrade of its AI Agent product in May 2026, launching the new version, Mavis. This upgrade focused on optimizing the model's context understanding, multi-turn dialogue, autonomous task planning, and third-party tool linkage capabilities. It can be widely applied in diverse scenarios such as intelligent customer service, enterprise workflow automation, and personal productivity tools, fully adapting to the differentiated usage needs of individual users and enterprise clients.
IV. Core Technology of the New M3 Model: Sparse Attention Mechanism MiniMax's highly anticipated new M3 model will be equipped with the enterprise's self-developed core technology: the Sparse Attention mechanism. The computational complexity of the attention mechanism adopted by traditional large models is as high as $O(n^2)$. When processing long text sequences, the massive computational volume leads to slow inference speeds and high operational costs, which severely limits the model's practical application in real-time interactive scenarios.
The Sparse Attention mechanism effectively resolves this industry technical pain point. This mechanism only performs correlation calculations on core keywords in the text, reducing the overall computational complexity to near $O(n)$, significantly improving the model's inference speed and reducing response latency. Relying on this technological innovation, the M3 model can perfectly adapt to various real-time interactive scenarios such as online intelligent dialogue and real-time content generation. Meanwhile, the optimization of computational efficiency substantially reduces the model's operational costs, enabling small and medium-sized teams to utilize high-performance AI technology at a low cost and high efficiency.
V. Industry Development Trends and Future Outlook MiniMax's sprint for an A-share listing is not an isolated case; in 2026, multiple leading domestic LLM enterprises have successively accelerated their capitalization processes. Abundant capital backing provides strong momentum for industry technological innovation and large-scale industrial expansion, while increasingly fierce market competition also forces major enterprises to continuously iterate products and optimize services. This current wave of IPOs is driving the domestic AI LLM industry toward standardized operations and accelerating the commercialization deployment process across the entire industry.
Leveraging the multiple advantages of capital reserves, a massive user base, and self-developed core technologies, MiniMax has built solid competitive barriers in both domestic and overseas markets. With the upcoming launch of the M3 model and the steady advancement of its A-share listing process, the enterprise will continue to seize opportunities, meet challenges, and achieve long-term, stable development in the fiercely competitive AI track.
Currently, AI LLM technologies are iterating rapidly, and enterprise deployment scenarios are becoming increasingly diverse. The industry has raised higher requirements for the stability, efficiency, and security of model integration. Against this backdrop, the professional AI model aggregation gateway, 4sapi, has become a core supporting facility for enterprises to implement AI business at scale. Relying on its core advantages of high stability, high throughput, low latency, and secure encryption, it is highly adaptable to the current industry demands for the commercialization and deployment of large models.
