When the model begins to grow the platform: the turning point of MiniMax

As one of the most watched AI model companies in Hong Kong stocks this year, MiniMax has delivered its first full-year performance report.

Along with the financial results, it also showcases its ambition to shift from a “model company” to an “AI platform-based company.” In the context of intensified global large model competition and capital markets gradually shifting from “storytelling” to “showing results,” this financial report clearly has more of a directional indicator significance.

Data shows that MiniMax achieved rapid revenue growth in 2025. Revenue from its AI-native products jumped from $21.805 million in 2024 to $53.075 million, a year-over-year increase of 143.4%. This is mainly due to the viral reputation of products like Heluo AI, Talkie/Xingye, and the significant increase in user willingness to pay globally.

Meanwhile, revenue from open platforms and enterprise services targeting B-end clients also surged by 197.8% to $25.963 million, demonstrating the scale effect of its model API in enterprise applications. The change in revenue structure indicates that MiniMax is moving away from relying solely on model call revenue, gradually forming a dual-driven business model with both consumer-end products and B-end services.

Against the backdrop of soaring revenue, MiniMax’s gross profit increased by 437.2%, reaching $20.079 million, with gross margin doubling from 12.2% to 25.4%.

Industry insiders believe that MiniMax has avoided the traditional internet-era “burning money for traffic” approach and has successfully paved a path driven by product strength for natural growth.

MiniMax COO Yuan Yeyi told Wall Street Insights that domestic major companies are probably most concerned with DAU, but what MiniMax cares about most is whether it can define a new generation of intelligence.

The product leader of MiniMax also stated, “AI is an era where production determines demand. What should we look at? The simplest is to see what models users are willing to pay for.”

Progress at the technical level is another main theme of this financial report.

In Q4 2025, MiniMax continuously launched models M2, M2.1, and M2-her, and in February 2026, further released version M2.5. Data shows that the average daily token consumption of the M2 series text models increased more than six times from December 2025 to February 2026, with coding-related consumption increasing over tenfold.

The explosive growth in token consumption indicates that improvements in model capabilities are driving genuine demand expansion.

Founder Yan Junjie stated that in the next year, the intelligence level of models will further leap, with programming entering L4 to L5 levels of intelligence, moving from “tools” to “colleague-level” collaboration; office scenarios will replicate the progress seen in programming last year; multimodal creation will achieve “direct output deliverables” of medium- and long-form content, even with real-time streaming output.

He believes that the combination of these three developments will trigger a large-scale explosion in intelligent supply and open an unprecedented window for application innovation. This judgment is not only a forecast of technological trends but also a declaration of the company’s strategic direction.

For MiniMax, last year was a decisive year for crossing from single-modal to full-modal integration.

The video model Heluo 2.3 released last year showed significant improvements in motion performance and image quality, reducing batch creation costs by up to 50%. By the end of the year, over 600 million videos had been generated. The Speech 2.6 voice model supported more than 40 languages, enabling ultra-low latency interactions, with over 200 million hours of speech generated. The Music 2.5 model achieved more complex emotional expression and longer output durations.

The collaboration of language, video, speech, and music modalities forms its “full-modal” technology matrix, providing underlying support for products like Heluo AI.

In addition to external product iterations, MiniMax is also conducting an organizational experiment. The company disclosed that its internal Agent interns now cover nearly 90% of employee roles, from programming to marketing, integrating AI collaboration. Early 2026, the company plans to productize internal capabilities and launch MiniMax Agent AI-native Workspace.

This “self-use first, then external sales” approach positions the company as an experimental ground for AI-native organizations, aiming to enhance organizational efficiency to feed back into product innovation. If successful, MiniMax could evolve from a model provider to a foundational infrastructure supplier for enterprise AI productivity platforms.

Overall, this annual report conveys three key signals. First, the path to commercialization is becoming clearer, with sustainable revenue growth. Second, gross margin has significantly improved, with technological efficiency beginning to translate into financial efficiency. Third, model capabilities are rapidly advancing and extending toward multimodal and Agent applications.

Of course, challenges remain.

High R&D investments will continue to suppress profit margins, and fluctuations in global computing costs and international competition pose ongoing risks. Profitability timelines are still uncertain. In 2025, MiniMax’s R&D expenditure reached $2.53 billion, a 33.8% increase year-over-year. Increased investments in cloud services and computing training led to an adjusted net loss of $250 million in 2025.

In the capital markets, as a star company in the Hong Kong AI sector, MiniMax’s valuation already reflects high growth expectations. Whether it can sustain its technological leadership and business expansion will be key factors influencing its stock performance.

From a long-term perspective, MiniMax stands at a critical inflection point. If the L4 to L5 programming capabilities truly materialize in 2026, enabling large-scale commercial use in office and multimodal content, the company’s transition from a “large model R&D provider” to a “platform company” will have a solid foundation. At that point, its role will no longer be limited to model calls but could become the infrastructure backbone of a new generation of productivity systems.

The 2025 performance report is a validation of capabilities; the execution in 2026 will determine whether MiniMax can truly complete this strategic upgrade.

Risk Warning and Disclaimer

Market risks exist; investments should be cautious. This article does not constitute personal investment advice and does not consider individual users’ specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions herein are suitable for their particular circumstances. Investment involves responsibility for the outcomes.