The “AI Six Tigers” were once favored by capital with their 100 billion parameter-level large models, but now they are facing product homogenization, commercialization difficulties and profitability problems. Recently, MiniMax has released a number of technical updates and news of its planned IPO in Hong Kong, and other members are also seeking listing or transformation.
As one of the “Six Little Tigers of AI”, MiniMax (Xiyu Technology) has just ended its “Technology Release Week”.
Since June 17, MiniMax has released a series of core technologies and product updates, covering multiple directions such as base models, video generation models, and multimodal agents, and systematically demonstrating its “model-multi-modal-application” integrated technical route.
At the same time as MiniMax was “crazy to launch”, the market reported the news of its planned IPO in Hong Kong, which also made MiniMax the second company to sprint to go public in the “AI Six Little Tigers” after Zhipu AI; However, the company has not responded to this at present.
Source: Screenshot of Weibo
Even if they are on the hottest track and have a celebrity resume, MiniMax or other “AI Six Tigers” companies are facing the same dilemma – product homogeneity, immature commercialization, and profitability difficulties.
01. The “Six Little Tigers” of the IPO
The “AI Six Little Tigers” (Spectrum, Dark Side of the Moon, Baichuan Intelligence, MiniMax, Step Stars, and Zero One Thousand Things) is a symbol of the domestic large model era in the past two years, referring to the earliest completion of $1 billion+ in financing that year, and all of which have self-developed 100 billion parameter-level large models, which are benchmarked with GPT-4, Llama and other large model startups in international benchmarks.
B-end product manager’s ability model and learning improvement
The first challenge faced by B-end product managers is how to correctly analyze and diagnose business problems. This is also the most difficult part, product design knowledge is basically not helpful for this part of the work, if you want to do a good job in business analysis and diagnosis, you must have a solid …
View details >
The six AI tigers can all be described as “capital darlings”. Behind MiniMax stand shareholders such as Sequoia China, Alibaba, Tencent, IDG Capital, and miHoYo, bringing together the top capital lineup in the industry and institutions. After receiving US$600 million in financing led by Alibaba in 2024, MiniMax has been valued at more than US$2.5 billion (equivalent to about 17.9 billion yuan); Recently, some media said that it is valued at about $3 billion.
Source: Screenshot of Tianyancha
In addition to being in the hottest AI track, MiniMax will also be sought after by capital, and it is also related to the beautiful resume of its founder and the company’s “anti-fighting” products.
This AI technology company, founded by Yan Junjie, former vice president of SenseTime, in 2021, for a long time, the company’s most “handy” product is an AI virtual companion application Talkie for overseas markets; In the first eight months of last year alone, Talkie’s global downloads quickly exceeded 10 million, surpassing CharacterAI to become the fourth most downloaded artificial intelligence application in the US market. According to media reports, user payment and advertising revenue from Talkie alone generated $70 million in revenue for MiniMax last year.
In order to avoid the risk of relying on a single product, in September 2024, MiniMax released the video generation model “Conch Video I2V-01” and launched Conch AI.
AIGC, as the first track to become popular after GPT, was actually considered a bit “outdated” when Conch AI was launched, not to mention that there were Kuaishou Keling AI in the video generation track and Byte Seeweed and PixelDance in the past.
Later, the “Moon Half Cat” IP video generated by Conch AI became popular, which also made Conch AI begin to go out of the circle. According to the “AI Product List Data”, since its launch, Conch AI has ranked first in video generation products in the global list for six consecutive months, surpassing Kuaishou’s Keling AI and Byte’s Dream AI.
Source: Screenshot of Xiaohongshu
Recently, MiniMax held a “technology release week”; For five consecutive days, five blockbuster products covering the underlying model, multimodal technology, general agent and other fields were thrown out.
As a support for the underlying basic capabilities, MiniMax released and open-sourced the self-developed MiniMax-M1 series models; In terms of supporting algorithms, MiniMax has launched a new CISPO reinforcement learning optimization algorithm; In the field of multi-modal generation, MiniMax launched a new generation of large video model Hailuo02, focusing on high-fidelity physics simulation and direct generation of 1080P quality video capabilities; HailuoVideoAgent uses natural language to drive global and full-process toolset calls to achieve “zero-threshold professional-grade video generation”; The voice model Speech02 has also been updated with VoiceDesign, which allows users to describe the timbre they want through natural language to achieve precise control of multiple dimensions.
Source: Screenshot of Weibo
However, in the same time period, MiniMax is not the only one with new model capabilities and agent products on the “AI Six Tigers”. On June 17, the dark side of the moon released a new open source code model Kimi-Dev-72B; Immediately afterwards, the dark side of the moon released its first Agent, Kimi-Researcher.
On July 2, Zhipu, which is impacting the IPO listing, officially released and open-source GLM-4.1V-Thinking, a large visual language model, and took the lead in open-source GLM-4.1V-9B-Thinking; At the same time, Zhipu also announced that Pudong Venture Capital Group and Zhangjiang Group have invested a total of 1 billion yuan in Zhipu, and recently completed the first delivery.
So far, some of the once high-spirited “AI Six Tigers” have canceled their businesses or shifted their focus, while others are still sticking to them and seeking listing financing.
02. Who is racing against?
Racing against time, racing against peers, and racing against capital is the current situation of the “AI Six Tigers”.
“The martial arts in the world are fast and not broken.” This is also very suitable for companies that use it to describe the AI track. For example, one of the most important reasons why MiniMax’s Talkie was out of the circle at that time was that it was launched fast enough.
In October 2022, a month before ChatGPT exploded, Talkie’s predecessor, Glow, an AI agent product that can have direct conversations, was launched; At that time, MiniMax was still testing a self-developed large language model with 30 billion parameters, and there were still many bugs to be solved.
Talkie, which was launched earlier, is not so much an AI application as a companion game application, so user stickiness and willingness to pay are very strong, and ToC charging has naturally become its commercial revenue model.
Product positioning, technology preemption, and market accuracy can be said to be the success factors of Talkie, but this is not easy to replicate by peers or even MiniMax itself.
For example, the conch Hailuo-02 model released by MiniMax has modified the membership system, with a one-year subscription fee of up to US$1,499.99 (RMB 10,788), and a fee to use the full video generation function; The skyrocketing cost has raised the threshold for video secondary creation, which has dissuaded many video creators at once, and has also caused large-scale user dissatisfaction and complaints.
Source: Screenshot of Conch
Also on a paid subscription basis, users of Talkie and Hailuo-02 have very different willingness to pay, in addition to the price factor, but also in the substitution of the product. “The homogenization of AI products is too high, and dialogue and image/video generation are the most common tracks, so we will see large models from various companies rushing to experience similar dialogue, search functions, or dialogue-derived functions, which leads to serious homogenization of many products.” AI industry practitioner Xiao Lin (pseudonym) said.
Serious homogenization will make users feel that there are many options, and once a product starts to charge, users can switch to other free products to continue to use. Therefore, the “price war” between peers has actually started before the product has begun to be realized.
Take conversational large models as an example, products such as Wen Xin Yiyan, Kimi, and IMA have already been launched, but why did large language models become popular again until the emergence of Deepseek? “Because it shows the process of thinking, in fact, this is not necessarily a technological leap, but it is an advancement in user experience, and users see the process of thinking about large models and have a more interactive experience.” An Agent practitioner commented.
Therefore, AI products must be “fast enough” or “distinctive”, and if they do not win the race against time, they must be differentiated compared with their opponents.
But whether it is a race against time or against an opponent, the final problem to be solved is profitability.
Even if several monetization routines are in front of you:Subscription system (ToC), data authorization (ToB), API or advertising (ToB), but the profitability of the “AI Six Tigers” is not particularly outstanding at present.
Product strength, differentiation, and profitability are the test questions that MiniMax and other “AI Six Tigers” must hand over their answers.
03. Who is the next super app
Looking back at the development of the Internet, the first generation of super applications such as BAT has been derived based on demand (search, social networking, e-commerce, etc.), and then differentiated new generation of super applications have been born in various segments, such as Pinduoduo and Xiaohongshu.
soWhat will super apps look like in the AI era?
OpenAI seems to have given a sample answer to this question, and it previously announced a five-level AGI roadmap, revealing the complete evolution path of artificial intelligence from a “chat tool” to a “strategic commander”. To put it simply, the first stage of AI is just a dialogue tool that can “speak human language” and can communicate without deliberate programming language; in the second stage, AI becomes a knowledgeable “doctor”, and the response and output are more accurate and faster; In the third stage, AI has become a super assistant, not only can “chat” but also “act”, and from here on, AI is not only a manipulated tool, but also can propose ideas and even manage a multinational company.
Source: Nuggets Technology Community
From this point of view, AI is currently in the transition period of the second and third stages. In other words, the “AI Six Little Tigers” have reached a critical point in the race against opponents and time, and the agents they come up with may be the key “players” to win this match point.
A serial entrepreneur in the AI track told “Fun Solution Business” that to put it simply, Chatbot is answering questions and Agent is executing instructions; For example, you need an agent to make a year-end summary PPT for you, this instruction can actually be broken down into several things – summary, summary, text refinement→ PPT template selection, illustration, design→ arrangement of page numbers→ PPT production; It will be natural for a human to disassemble this task, and it can even be written where to go first, but the Agent needs multi-modal cooperation to execute, and the disassembly, division of labor, thinking, and completion of the task in the middle are technical problems, which will directly affect the completion of the task.
“Why are many people doing general AGI in China, because there is no ‘killerAPP’ in general AGI, forming a monopoly situation like OpenAI, so from a market perspective, this blue ocean is very large; “On the other hand, in terms of technical difficulty, a general AGI plus the knowledge base and private database of the vertical track can be transformed into a vertical agent, so unless the enterprise has anchored a vertical scenario to make a sufficient foundation for commercialization, Otherwise, premature verticality is actually erasing other possibilities for itself. ”
Whether it is MiniMax’s Agent product or Kimi-Researcher, which Kimi recently tested in a small area, it is also a product that meets the concept of general AGI; This can also be seen that Liu Xiaohu’s idea of wanting to eat this bite of cake first is very similar.
It can also be seen from the evolution of chat to Agent that AI has gradually upgraded from the “hardware” of volume, researchers, and R&D expenses to the “soft power” of volume positioning, volume market, and volume user experience. And on this track, the seed player of the next super application may be born.
