--- title: "The world's first performance report on large models! MiniMax predicts three major super PMFs for 2026, and AI platform companies are on their way" type: "News" locale: "en" url: "https://longbridge.com/en/news/277558187.md" description: "MiniMax released the world's first annual report for large model companies, showing that revenue will reach USD 79.04 million in 2025, a year-on-year increase of 158.9%. Gross profit increased by 437% year-on-year, and the loss rate significantly narrowed. In 2026, the ARR is expected to exceed USD 150 million, with 70% of revenue coming from international markets. MiniMax's revenue mainly comes from AI-native products and open platforms, with over 236 million users, demonstrating strong commercialization potential" datetime: "2026-03-03T03:51:09.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/277558187.md) - [en](https://longbridge.com/en/news/277558187.md) - [zh-HK](https://longbridge.com/zh-HK/news/277558187.md) --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/277558187.md) | [繁體中文](https://longbridge.com/zh-HK/news/277558187.md) # The world's first performance report on large models! MiniMax predicts three major super PMFs for 2026, and AI platform companies are on their way Here it comes! Global large model companies finally have a truly quantifiable financial sample. Just 52 days after listing on the Hong Kong Stock Exchange, MiniMax has delivered its first annual report since the IPO: By February 2026, the ARR will exceed $150 million, with 2025 revenue growing by 158.9% year-on-year, gross profit soaring by 437% year-on-year, and a significant narrowing of the loss rate... But the more critical significance lies in: as the world's first annual report from a large model company, it provides a valuable window for the market to gain insights into "how large models are commercialized," and serves as an important barometer for measuring whether Chinese AI companies can succeed in global competition. Starting from this point, we can gain insights not only into MiniMax's next steps but also into the evolutionary trends of the entire AI large model industry. ## **How does the world's first annual report from a large model company perform?** In 2025, which is the second year of MiniMax's true commercialization, the company's total revenue for the year reached $79.04 million, a **staggering increase of 158.9%** year-on-year, with **over 70%** of the revenue coming from international markets. The company's adjusted net loss for the past year was $250 million, with the **net loss rate** significantly narrowed. Simply put: the money earned is increasing, while the losses are decreasing. MiniMax's revenue achieved exponential growth, with the dual-driven situation of AI native products and open platforms becoming more apparent. Specifically, MiniMax's revenue sources can be divided into two main categories: AI native products aimed at the **C-end** and open platforms and other AI-based enterprise services aimed at the **B-end**. Revenue from AI native products refers to subscription income from applications such as **MiniMax, MiniMax Voice, Conch AI, and Xingye**. By the end of 2025, MiniMax had served over **236 million** users across more than 200 countries and regions. This user scale is quite impressive in the global internet product landscape. Due to the continuous advancement of product commercialization, as well as increased user engagement and willingness to pay, in 2025, this part of the business contributed $53.08 million in revenue to MiniMax, a year-on-year increase of 143%, accounting for 67.2% of total revenue. For enterprise and developer open platforms and other AI-based enterprise services, which charge based on usage, **last year brought MiniMax $25.963 million in revenue, a year-on-year increase of 197.8%**. During the conference call, MiniMax founder Yan Junjie also revealed the latest news—product growth in text model-driven productivity scenarios has become more pronounced this year **For example, the average daily token consumption of the M2 series text models in February 2026 has increased to more than 6 times that of December 2025, with token consumption from CodingPlan growing more than 10 times.** In February of this year, the number of new registered users in the open platform products aimed at enterprise clients and individual developers has **reached more than 4 times that of December 2025**. Currently, MiniMax has **214,000 enterprise clients and developers** from over 100 countries and regions, and more than 50% of the revenue from the development platform comes from overseas markets. **This "C-end + B-end" dual-driven business model provides MiniMax with stable growth and predictable recurring revenue, significantly enhancing its profitability.** In 2025, MiniMax's gross profit was $20.08 million, a staggering increase of 437% year-on-year, and significantly higher than the revenue growth rate. During the same period, the growth rate of the company's **gross margin** was also impressive, having improved from -24.7% in 2023 to 12.2% in 2024, and then increased by 13.2 percentage points to **25.4%** in 2025. The improvement in gross profit levels is mainly due to the enhancement of model and system efficiency, as well as the optimization of infrastructure configuration. Combined with a significantly narrowed net loss rate, from a profit perspective, MiniMax is one step closer to a sustainable commercialization loop. To achieve such capabilities and results, initial investment and groundwork are essential. In 2025, MiniMax's R&D expenditure was $250 million, a year-on-year increase of 33.8%. For an AI large model company, in addition to intuitive numbers, R&D efficiency is another important metric to consider: The proportion of MiniMax's R&D expenditure to total revenue has decreased from 619% in 2024 to 320% in 2025, indicating that R&D efficiency is gradually improving. At the same time, the term "efficiency" is also reflected in MiniMax's cash reserves. As of the end of 2025, the company's cash reserves amounted to **$1.05 billion** (including cash and cash equivalents, restricted cash, and time deposits, etc.), up from $880 million at the end of 2024. Moreover, after going public in Hong Kong, the capital market has voted with its feet, and MiniMax's stock price has risen, further enhancing its cash "ammunition depot." Overall, looking at this performance, MiniMax has entered an accelerated running phase. So what kind of technology and products support such explosive growth? ## **Intensive Product Iteration, Accelerated Commercialization** For MiniMax, 2025 is a year of comprehensive technological advancement and accelerated commercialization. In this year, MiniMax has built a full-modal R&D capability, with globally competitive models in key modalities such as language, video, voice, and music. In the 108 days from the fourth quarter of 2025 to the beginning of this year, MiniMax continuously completed updates for three generations of language model products: M2, M2.1, and M2.5, demonstrating an industry-leading model iteration speed. Especially in the second half of the year, MiniMax completely opened the curtain for a product explosion. Looking back along the timeline, it is not difficult to find that MiniMax has been in a rapid iteration mode since October of last year. When M2 was released, the R&D team reverted to the traditional full attention mechanism for the stability of real tasks, and **broke the impossible triangle of intelligent operational speed and computing cost**. In practical scenarios, this model won first place in the AI-Trader simulated A-share competition at the University of Hong Kong, earning nearly three thousand yuan with an initial capital of 100,000 yuan in 20 days. By December, MiniMax passed the Hong Kong Stock Exchange hearing and subsequently launched the flagship Coding&Agent model M2.1. M2.1 **achieved global SOTA in multilingual programming and addressed a major shortcoming of Vibe Coding**, mastering both front-end and back-end development specifications. Later, after the introduction of OpenClaw, the author of OpenClaw praised M2.1 as the (then) best open-source model. Less than two months after M2.1 went live, around the time of the Spring Festival, the M2.5 model was also launched, **still focusing on code and agent capabilities, but becoming lighter, faster, and more cost-effective**. Its inference speed can reach 100 TPS, capable of complete development for front-end and back-end databases, and running as an agent costs only 1 dollar per hour. Today, Notion co-founder Akshay Kothari announced that **Notion Custom Agents has officially introduced the first open-source weight model MiniMax M2.5**, which also means that MiniMax M2.5 is continuously penetrating the core scenarios of international mainstream productivity tools. During this period of continuous model upgrades, **MiniMax has also updated the "scaffolding" of model agent capabilities—the MiniMax Agent platform—twice** In January of this year, MiniMax Agent launched version 2.0, introducing a desktop version that can **deeply integrate into the local work environment**, directly reading files from the computer to get the job done. Additionally, the simultaneously launched Experts system allows users to **create exclusive vertical domain experts through unique knowledge and workflows in various task scenarios**. After the Spring Festival, the Agent platform welcomed another update, further strengthening community co-construction of the Experts system, with **the number of applications exceeding ten thousand**. Returning to MiniMax's product iteration path, the MiniMax Agent platform followed the OpenClaw trend and introduced the **MaxClaw** model. Users do not need to worry about complex underlying configurations; cloud deployment can be completed with one click on the webpage, allowing those with zero foundation to enjoy lobster. ## **Embracing Three Super PMFs for 2026, Becoming an AI Platform Company** Today, the iteration speed in the large model track is accelerating, which also means competition is becoming more intense, and the market is changing rapidly. The next generation of AI companies that will win the future cannot be pure technology providers; they must be organizations with strong comprehensive capabilities in technology, products, commercialization, and more. The future winners **must first be action-oriented companies that have moved away from a test-taker mindset**. Current competition is no longer about who scores higher; it’s about who can achieve results in a real chaotic business environment. Take MiniMax's latest M2.5 as an example; it possesses true full-stack delivery capabilities that encompass backend logic and database design. Combined with MiniMax Agent to bridge the cloud and local computer environments, allowing AI to actively read local files and digest implicit business experiences, large models truly become knowledgeable long-term "digital partners." What determines the life and death of a large model company is also technical judgment, market intuition, and evolutionary speed. In the words of MiniMax CEO Yan Junjie during the earnings call, it is about having the ability to define a new intelligent paradigm and preparing in advance for it **.** Starting from the second half of 2025, MiniMax has been actively preparing for what they believe will emerge this year as several super PMFs: - In the programming field: AI programming will reach L4 to L5 levels of intelligence, with AI transitioning from a "tool" to "colleague-level" collaboration. L4 can be understood as the ability to propose innovative solutions to engineering and complex algorithm problems; L5 requires multiple intelligent agents to be effectively organized and collaborate like humans; - In the office field: The workspace scenarios for various professions represent a broader coverage and larger market space, which will be more complex than programming, but MiniMax believes that office scenarios will replicate last year's progress speed in the programming field; - In multi-modal creation: AI creation tools will be able to generate "deliverable" medium to long content, even producing new forms of streaming and real-time output. For example, MiniMax has already upgraded the Media Agent in the HaiLuo AI product to support all-modal, all-capable creation, allowing for one-click production This means that new technological challenges are about to emerge, and a larger scale of intelligent supply will explode, with the application layer welcoming a huge window of innovation. Demand for AI companies like MiniMax, which are centered around multimodal models, will be amplified, and **the token scale is likely to see an increase of 1-2 orders of magnitude**. To this end, the MiniMax organization is prepared and is developing the M3 and Hailuo3 series models specifically designed to address these challenges. In the face of such opportunities, MiniMax will fully upgrade from a **large model company** to an **“AI platform company,”** with **intelligent density and model throughput** as core metrics, driving AI to become a new generation of global production infrastructure. In MiniMax's view, when the boundaries of intelligence are broken, a large number of new scenarios and new users will emerge, further forming new ecosystems and new commercialization dividends. Companies that can define this boundary of intelligence and enjoy the dividends brought by this boundary in terms of products and business are what they define as “AI platform companies.” In simpler terms, the value of an AI platform company in the era of AI can be roughly estimated as providing **intelligent density × token throughput**. When both are strong enough, the value of the platform will naturally manifest. In MiniMax's roadmap, M3 and Hailuo 3 are the ultimate practitioners of this estimation formula—research and development of the M3 and Hailuo 3 series models will continuously optimize reasoning architecture and computational efficiency. It is foreseeable that the boundaries of intelligence are being redefined, and 2026 will be a critical watershed for AI to transition from the “tool era” to the “ecosystem era.” Risk Warning and Disclaimer The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investing based on this is at one's own risk ### Related Stocks - [MINIMAX-WP (00100.HK)](https://longbridge.com/en/quote/00100.HK.md)