The "Six AI Tigers" have reached the next critical juncture.
As one of the "Six AI Tigers," MiniMax recently concluded its "Technology Release Week."
Since June 17th, MiniMax has unveiled a series of core technologies and product updates, covering multiple areas such as base models, video generation models, and multimodal intelligent agents, systematically demonstrating its integrated technology roadmap of "Model - Multimodality - Application."
While MiniMax was "rapidly launching new products," news spread in the market that it plans to go public in Hong Kong. This makes MiniMax the second company among the "Six AI Tigers" to be reported as aiming for an IPO after Zhipu AI. However, the company has not responded to this news yet.
Image source: Screenshot from Weibo
Even though they are in the hottest current field and have star - studded resumes, MiniMax and other companies among the "Six AI Tigers" are facing the same dilemmas - product homogenization, immature commercialization, and profit difficulties.
01. The "Six Tigers" Rushing for IPO
The "Six AI Tigers" (Zhipu, DarkSide of the Moon, Baichuan Intelligence, MiniMax, Jieyue Xingchen, and Lingyi Wanwu) are a symbol of the domestic large - model era in the past two years. They refer to the large - model startup companies that were the first to complete over $1 billion in financing, each having self - developed large models with hundreds of billions of parameters, and competing with models like GPT - 4 and Llama in international benchmark tests.
The Six AI Tigers are all "favorites of capital." Behind MiniMax stand shareholders such as Sequoia China, Alibaba, Tencent, IDG Capital, and Mihoyo, bringing together top - tier capital from both industries and institutions. After receiving a $600 million financing led by Alibaba in 2024, MiniMax's valuation has exceeded $2.5 billion (equivalent to about 17.9 billion RMB); recently, some media reported its valuation to be about $3 billion.
Image source: Screenshot from Tianyancha
Besides being in the hottest AI field, MiniMax's popularity among capital is also related to its founder's excellent resume and the company's "robust" products.
This AI technology company, founded in 2021 by Yan Junjie, the former vice - president of SenseTime, had for a long time only one "outstanding" product - Talkie, an AI virtual companion app targeting the overseas market. In the first eight months of last year alone, Talkie's global downloads quickly exceeded 10 million, surpassing Character AI and becoming the fourth most - downloaded AI app in the US market. Some media reported that last year, user payments and advertising revenue from Talkie alone generated $70 million in revenue for MiniMax.
To avoid the risk of relying on a single product, in September 2024, MiniMax released the video generation model "Hailuo Video I2V - 01" and launched Hailuo AI.
As the first popular field after GPT, AIGC was already considered a bit "outdated" when Hailuo AI was launched. Moreover, in the video generation field, there were already Kuaishou's Keling AI, and ByteDance's Seeweed and PixelDance. As a product of a startup, Hailuo AI was not highly expected by the market at that time.
Later, the "Yuebanmao" IP videos generated by Hailuo AI became extremely popular, and Hailuo AI started to gain popularity. According to "AI Product List Data," since its launch, Hailuo AI has ranked first among video generation products on the global list for six consecutive months, surpassing Kuaishou's Keling AI and ByteDance's Jimeng AI.
Image source: Screenshot from Xiaohongshu
Recently, MiniMax held a "Technology Release Week." For five consecutive days, it launched five major products covering areas such as underlying models, multimodal technologies, and general intelligent agents.
As the support for underlying basic capabilities, MiniMax released and open - sourced its self - developed MiniMax - M1 series of models. In terms of supporting algorithms, MiniMax introduced a new CISPO reinforcement learning optimization algorithm. In the field of multimodal generation, MiniMax launched a new - generation large video model, Hailuo 02, which features high - fidelity physical simulation and the ability to directly generate 1080P - quality videos. Hailuo Video Agent enables "professional - level video generation with zero threshold" by driving the global and full - process toolset through natural language. The speech model, Speech 02, also had a function update. With Voice Design, users can describe the timbre they have in mind through natural language to achieve precise control in multiple dimensions.
Image source: Screenshot from Weibo
However, during the same period, MiniMax was not the only one among the "Six AI Tigers" to launch new models and Agent products. On June 17th, DarkSide of the Moon released a new open - source large code model, Kimi - Dev - 72B. Immediately afterwards, it released its first Agent, Kimi - Researcher.
On July 2nd, Zhipu, which is aiming for an IPO, officially released and open - sourced the visual - language large model GLM - 4.1V - Thinking and was the first to open - source GLM - 4.1V - 9B - Thinking. At the same time, Zhipu announced a strategic investment of 1 billion RMB from Pudong Venture Capital Group and Zhangjiang Group, and the first payment was completed recently.
So far, among the once - high - flying "Six AI Tigers," some have cut back on business or shifted their focus, while others are still holding on and seeking IPO financing.
02. Who Are They Racing Against?
Racing against time, against peers, and against capital - this is the current situation of the "Six AI Tigers."
"In martial arts, speed is the key to victory." This also applies well to companies in the AI field. For example, one of the most important reasons why MiniMax's Talkie became popular was its fast launch.
In October 2022, one month before ChatGPT became extremely popular, Glow, the predecessor of Talkie, an AI intelligent agent product that allowed direct conversations, was already launched. At that time, MiniMax was still in the internal testing phase of its self - developed large language model with 30 billion parameters, and there were still many bugs to be fixed.
Talkie, which was launched early, was more like a companion game app than an AI app, so users had strong stickiness and high willingness to pay. Naturally, ToC charging became its commercial revenue model.
Product positioning, technological leadership, and finding the right market can be said to be the success factors of Talkie, but these are difficult to replicate even by MiniMax itself or its peers.
For example, MiniMax's Hailuo 02 model revised its membership system, with the highest annual subscription fee reaching $1499.99 (10788 RMB), and users need to pay to use the full video generation function. The sharp increase in cost raised the threshold for video re - creation, which discouraged many video creators and caused widespread user dissatisfaction and complaints.
Image source: Screenshot from Hailuo
For the same paid subscription system, users' willingness to pay for Talkie and Hailuo 02 is very different. Besides price, it also lies in the product's substitutability. "The current AI products are highly homogeneous. Conversation, image/video generation are the most common fields. So we can see that large models from various companies are rushing to launch similar conversation and search functions or functions derived from conversations, which leads to serious homogenization of many products," said Xiao Lin (a pseudonym), an AI industry practitioner.
Severe homogenization makes users feel that they have many options. Once a product starts charging, users can easily switch to other free products. So the "price war" among peers has actually started before the products start to generate revenue.
Take the conversation - based large models as an example. Products like Wenxin Yiyan, Kimi, and IMA have been launched for a long time, but why did large language models become popular again only after the emergence of Deepseek? "Because it shows the thinking process. In fact, this may not be a technological leap, but it is an improvement in user experience. Users can see the thinking process of the large model and have more interactive experiences," an Agent practitioner commented.
Image source: Canstock
So AI products need to be "fast enough" or "distinctive." If they can't win the race against time, they need to be different from their competitors.
But whether racing against time or against competitors, the ultimate problem to be solved is the same - profitability.
Even though several monetization methods are available: subscription system (ToC), data authorization (ToB), API, or advertising (ToB), currently, the profitability of the "Six AI Tigers" is not particularly outstanding.
Product strength, differentiation, and profitability are the questions that MiniMax and other "Six AI Tigers" must answer.
03. Who Will Be the Next Super App?
Looking back at the development of the Internet, starting from user needs (search, social, e - commerce, etc.), the first - generation super apps like BAT emerged, and then in various niche fields, differentiated new - generation super apps such as Pinduoduo and Xiaohongshu were born.
So, what will super apps look like in the AI era?
OpenAI seems to have provided a sample answer to this question. It previously announced a five - level roadmap for AGI, revealing the complete evolution path of AI from a "chat tool" to a "strategic commander." Simply put, in the first stage, AI is just a chat tool that can "speak human language," allowing communication without the need for a specific programming language. In the second stage, AI becomes a knowledgeable "doctor," with more accurate and faster responses and outputs. In the third stage, AI becomes a super assistant, capable of not only "chatting" but also "acting." From this stage on, AI is no longer just a tool to be controlled but can also come up with ideas and even manage a multinational company.
Image source: Juejin Technology Community
From this perspective, AI is currently in the transition period between the second and third stages. That is to say, the "Six AI Tigers" have reached a critical point in the race against opponents and time, and the Agents launched by each company may be the key "players" to win this race.
A serial entrepreneur in the AI field told "Qujie Business" that simply put, a Chatbot answers questions, while an Agent executes instructions. For example, if you need an Agent to create a year - end summary PPT, this instruction can be broken down into several tasks - summarizing and generalizing the work of the year, extracting key points → selecting a PPT template, adding illustrations, and designing → arranging page numbers → creating the PPT. A human can break down this task naturally and may even write as they go. However, for an Agent to execute this task, it requires multimodal cooperation. The task breakdown, division of labor, thinking, and completion in this process are technical challenges that will directly affect the task completion.
"Why do many people in China focus on general AGI? Because there is no 'killer app' for general AGI yet, and there is no monopoly situation like that of OpenAI. So from a market perspective, this is a huge blue ocean," he said. "On the other hand, in terms of technical difficulty, a general AGI combined with a knowledge base and private database in a vertical field can be transformed into a vertical Agent. So unless a company has established a solid commercial foundation in a vertical scenario, premature verticalization may also eliminate other possibilities."
Both MiniMax's Agent products and Kimi - Researcher, which Kimi recently tested in a small - scale internal test, are products that conform to the concept of general AGI. This shows that the Six Tigers have similar ideas about seizing this opportunity.
From the evolution from chat to Agent, we can see that AI has gradually upgraded from competing in "hardware" such as data, researchers, and R & D expenses to also competing in "soft power" such as positioning, market, and user experience while having strong hardware capabilities. In this field, there may be a candidate for the next super app.
This article is from the WeChat official account "Qujie Business", written by Liu Liang, and published by 36Kr with authorization.