MiniMax stands out as a unique case among the AI Little Tigers, having charted a distinct course from the outset. In the realm of general-purpose large models, MiniMax adheres to a ‘model-product integration’ philosophy, ensuring that all foundational models directly serve specific product applications. While Kimi primarily targets the domestic market, MiniMax strategically prioritizes overseas expansion. Amidst the relentless pursuit of user retention, growth, and monetization, MiniMax has reportedly generated an impressive $70 million in annual revenue from its popular overseas product, ‘Talkie.’
‘A superior model can lead to a superior application, but a superior application and more users do not necessarily lead to a superior model.’ This statement encapsulates MiniMax’s core belief.
Despite 闫俊杰’s early recognition of the importance of technical investment, predating DeepSeek’s surge in popularity, MiniMax faces significant uncertainties. These include navigating the volatile overseas market landscape and confronting fierce competition from domestic tech giants. Furthermore, its initial lack of focus on the B-end (business-to-business) market requires immediate attention.
As of this year, the AI Little Tigers, with the exception of 智谱, appear to have encountered challenges in securing further funding. Can MiniMax’s renewed emphasis on model development, product innovation, and monetization enable it to break through the competitive landscape?
From Model-Product Integration to Separation
MiniMax’s models are fundamentally different from others.
According to official documentation, the MiniMax-01 series models deviate from traditional Transformer architectures, pioneering the large-scale implementation of linear attention mechanisms. To illustrate this concept, consider the traditional Transformer as a ‘brute-force’ method, akin to reading a text word by word and meticulously taking notes. Linear attention, conversely, employs a more efficient approach, similar to group discussions culminating in a final consensus.
In essence, MiniMax’s ‘underlying innovation’ lies in modifying the fundamental architecture of large models. This transformation, akin to upgrading from a bus to a light rail system, significantly enhances efficiency and reduces costs.
For an extended period, MiniMax championed the concept of ‘model-product integration,’ emphasizing that models should be designed to serve specific applications. In practical terms, each MiniMax model is paired with a corresponding AI application product. For example, the text model powers the MiniMax assistant, the video model drives 海螺AI, and 星野 and Talkie represent comprehensive technological showcases.
This technology-to-product logic has been evident in MiniMax’s multimodal approach from an early stage. Industry experts suggest that MiniMax has excelled in identifying viable application scenarios, enabling individual technologies such as text-to-image, text-to-video, and speech to effectively contribute to C-end (consumer-facing) applications, resulting in substantial commercial success.
‘While MiniMax’s revenue may not match the scale of the To B-focused 智谱, its quality is higher, and positive user feedback will drive steady growth,’ according to one industry insider.
According to 闫俊杰, the shift in strategy occurred around March or April of last year. At that time, he reassessed the relationship between users and technology, concluding that ‘the improvement of intelligence levels does not depend so much on many users.’ He now believes it is more important to focus on the core technology to drive growth.
闫俊杰 stated in an interview, ‘We are very clear that we are a technology-driven company. It is not just a slogan, but in essence, who has the final say when there is a conflict.’ This underscores the company’s unwavering commitment to technology. MiniMax’s strategy heavily relies on technology advancements as a core driver.
Consequently, the model-product integration approach was abandoned. Technology and products were separated, with technology focusing on continuously raising the upper limit and products focusing on better meeting user needs. The company aims to excel in both technological innovation and product user experience.
To some extent, 闫俊杰 rejects the current AI application logic of using investment-driven marketing to drive growth. It is understood that MiniMax has also focused on promoting Talkie and 海螺AI overseas. According to DataEye data, 星野 once became the number one on the daily list of AI App advertising materials in November and December last year, with the amount of materials increasing several times compared with the previous period.
Following DeepSeek, MiniMax released and open-sourced the MiniMax-01 series models, which include two models: the basic language model MiniMax-Text-01 and the visual multimodal model MiniMax-VL-01. In addition, it is betting on long-context and Agent technology in terms of technology direction.
HaiLuo AI, The Rising Star
In terms of products, MiniMax has taken two actions. The first is to implement a focus strategy and unify the brand image for external output. The most well-known ‘HaiLuo’ was reserved for the video generation business, and the original AI assistant ‘HaiLuo AI’ was renamed MiniMax. The brand seeks to simplify its product offerings and strengthen its market position.
Another new trend is also related to AI video. There are reports that MiniMax will acquire Shenzhen AI video generation startup 鹿影科技, and the two parties have currently finalized the acquisition intention. It is reported that the core product of this startup is YoYo, a two-dimensional animation AI video generation platform. MiniMax’s acquisition may be to further expand the user base, from film and television creators and AI creators to the two-dimensional community. By acquiring a platform specializing in 2D animation, MiniMax aims to diversify its user base and capture a larger market segment.
All signs indicate that after 星野 and Talkie, HaiLuo AI video products have become MiniMax’s new trump card. The strategic shift suggests that MiniMax views AI video as a critical area for future growth.
The reason for this may be related to the delisting behavior at the end of last year. According to data from 点点数据, Talkie was removed from the US App Store on December 14, 2024, and it had already been removed from the Japanese App Store on November 30. The delisting event likely influenced MiniMax’s strategic decision to emphasize HaiLuo AI as an alternative revenue stream.
As of press time, when we searched for ‘Talkie’ in the US App Store, we only found ‘Talkie Lab’ and did not find the real application. However, according to App Annie data, Talkie applications are still generating new downloads and payment fees. While the application may not be directly available in some app stores, it still contributes to MiniMax’s revenue.
Currently, MiniMax’s main income comes from Talkie and 星野, the advertising and recharge subscription fees of the two main products. App Annie data shows that from April 1, 2023 to the end of March 2025, Talkie’s total App Store revenue was US$3.218 million (approximately RMB 23.519 million); from September 2023 to the end of March 2025, 星野’s (only launched in China) total revenue was US$244,000 (approximately RMB 178,000). Talkie’s revenue significantly outweighs that of 星野, indicating its substantial contribution to MiniMax’s financial performance.
Coupled with advertising revenue, Talkie’s importance to MiniMax is self-evident. 光子星球 found through整理数据 that Talkie’s top five countries in terms of downloads on the App Store are the United States (28%), Germany (6%), the United Kingdom (5%), Canada (3%), and Australia (2%). The data reveals the key markets driving Talkie’s download numbers.
Under normal circumstances, we believe that the country with a larger proportion of downloads should naturally contribute more revenue. According to the figure above, it means that 56% of other regions should generate roughly the same proportion of revenue. But the opposite is true. The United States, which only generates 28% of application downloads, ultimately contributed 68% of the revenue. Data shows that the more developed countries and regions are, the stronger the willingness to pay. This explains why MiniMax attaches so much importance to the US market and why it launched a frantic investment campaign after being removed from the shelves. The user base in developed countries demonstrates a higher propensity for in-app purchases, making them a lucrative market for MiniMax.
AI companion products and Chinese overseas companies are labeled with two labels, which exacerbates the risk level of Talkie. Either it will be taken off the shelves by public opinion, or it will become a sacrifice in the game between China and the United States. The association with AI companion products and being a Chinese overseas company introduces regulatory and geopolitical risks for Talkie.
To prevent this from happening, MiniMax put its treasure on Hailuo AI video, which has less risk and is linked to productivity. By focusing on HaiLuo AI video, MiniMax aims to mitigate the risks associated with Talkie and explore a more sustainable and compliant business model.
From the current point of view, the market feedback of Hailuo AI video is positive, and its growth momentum is very rapid. App Annie shows that Hailuo AI was gradually launched on the global App Store on February 19 this year. As of April 1, in just forty-two days, its total global downloads reached 386,000, and its total App Store revenue was US$29,000, and its potential should not be underestimated. The quick market acceptance and initial revenue figures of Hailuo AI video suggest a promising future for the product.
Taking on the Challenge of B-End Customers?
At this point, MiniMax has formed a product matrix of Talkie, 星野 and Hailuo AI, and its revenue from globalization and C-end AI applications still has the opportunity to achieve steady growth. In this context, the shortcomings of its B-end business are becoming more and more obvious.
During the turmoil period, MiniMax’s To B business is undergoing a dual test of people and business. The company faces both internal personnel changes and business model challenges.
Smart Surging News said that MiniMax co-founder and vice president Wei Wei has left his job. He was previously mainly responsible for the commercialization of the To B direction and had held important positions in Tencent Cloud and Baidu Smart Cloud. The departure of a key leader in the B-end commercialization effort presents a setback for the company.
Related people told us that MiniMax’s enterprise business is basically a natural traffic model. Customers come to inquire and do it. It will not actively nurture a huge To B sales team, and mainly sell API models to make deals. MiniMax’s reliance on inbound inquiries and API sales in the B-end market suggests a passive approach that may limit its growth potential.
Another customer who contacted MiniMax’s enterprise business once euphemistically pointed out in an interview with us that MiniMax, which is obsessed with technology, exudes the temperament of a science and engineering man and sometimes seems to lack a bit of ‘EQ.’ The perception of MiniMax as overly focused on technology and lacking interpersonal skills in the B-end market could hinder its ability to establish strong customer relationships.
Before DeepSeek, industry customers highly recognized MiniMax’s voice large model capabilities. Among them, AI toy manufacturer Haivivi BubblePal, Yuewen Qidian audiobook, and Gaotu Education are all its customers, and the main application scenarios include AI dialogue, AI teaching, and AI storytelling. MiniMax had previous successes in deploying its voice models for various applications, but it needs to innovate continuously.
After DeepSeek, MiniMax turned its attention to the field of smart bodies and smart hardware, and announced the establishment of the ‘MiniMax Smart Hardware Industry Innovation Alliance’ with smart home, wearable equipment, smart cockpit and many other smart hardware companies. The formation of an alliance with smart hardware companies reflects MiniMax’s intent to expand into new markets, but the effectiveness of this strategy remains to be seen.
It is not difficult to see from the above that MiniMax’s style in the B-end business tends to be light delivery. The advantage is that the cycle is short and the risk is small, but this indirectly stifles the possibility of it being deeply rooted in the B-end field. The B-end large model has always been a tough ‘hard bone’ to crack. Each subdivided industry is like a complicated tree with intertwined roots. The more you dig deeper, the more difficult it becomes. 智谱 relies on university resources to not only do Xinchuang but also do a full set of solutions to capture high-customer-price customers, and only then can it barely squeeze into the ranks of cloud vendors. The light delivery model employed by MiniMax may not be sufficient to penetrate the complex and demanding B-end market.
Not only MiniMax, but many AI companies are also locking in ecological positions through alliances. However, this form of alliance is precarious and mostly presents a relationship of mutual platform. Cooperative companies use it and leave. If it does not infringe on their interests, they are happy to endorse it a few more times. In the end, the depth of cooperation between the two parties can determine the effect of rooting in the industry. The effectiveness of alliances in securing long-term partnerships depends on the depth of collaboration between the parties involved.
Enterprise B-end business is often not as straightforward as technology. It is full of games and trade-offs. MiniMax, which wants to crack the hard bones of To B, needs a little determination and time to settle down. Overcoming the challenges of the B-end market requires MiniMax to adopt a more strategic and patient approach.