MiniMax: Sweetness and Hidden Worries Under the Revenue of 70 Million USD | Focus Analysis

By ZHOU Xinyu

Edited by SU Jianxun

I believe everyone has already heard the latest news about MiniMax: foreign media outlets have revealed that MiniMax's projected annual revenue stands at 70 million USD. This delivers a powerful counterargument to the claim that "large model companies cannot find PMF (Product-Market Fit)".

We have learned that this figure is actually a conservative estimate.

An insider disclosed to *Intelligent Emergence* that MiniMax's actual annual revenue is most likely to exceed 70 million USD. The primary source of this revenue comes from Talkie, MiniMax's overseas consumer-facing AI application.

The Talkie interface.

This character companion AI application, launched in June 2023, has rapidly created a growth miracle overseas by leveraging AI's comprehension and generation capabilities, as well as the booming momentum of the AI industry:

According to Sensor Tower's monitoring data, since its release, Talkie has consistently ranked among the top three most downloaded companion AI applications worldwide, alongside two well-known American unicorns, Character.AI and Replika.

It is understood that the rapid growth of users has also brought considerable advertising revenue to Talkie. Of course, as a consumer-facing application, Talkie's monetization model for users mainly focuses on membership subscriptions and in-app purchases to unlock features such as gacha draws.

Talkie membership subscription interface. Source: Internet

Talkie's success has sparked a wave of character companion AI applications in China. However, neither ByteDance's "Catbox" nor "Customize Ta", which was once a hit in anime communities, has managed to replicate Talkie's global growth trajectory.

The fact that Talkie's success is so hard to replicate has left MiniMax in a stalemate as it attempts to expand its business using the same methodology.

Today, MiniMax's valuation has exceeded 200 billion yuan — to strengthen its business moat and risk resilience, the company needs more products like Talkie.

Multiple industry stakeholders told *Intelligent Emergence* that whether it's Talkie's domestic version "Stellar" that MiniMax is heavily promoting, or its productivity tool "Conch AI", MiniMax wants to replicate the same strategy on both products: adding companion and social elements.

However, even the creators of Talkie have failed to build another Talkie.

Why Talkie Took Off: Early Launch, Polished Execution

Talkie's explosive growth is undoubtedly of high reference value for domestic entrepreneurs who are currently working tirelessly to find AI PMF (Product-Market Fit).

As a company that only rose to prominence in the large model wave, how did MiniMax create a product with tens of millions of dollars in annual revenue within a single year? After interviewing multiple industry professionals, *Intelligent Emergence* summarizes the answer as: timing of launch, product positioning, and rapid refinement and iteration.

Being early and being fast are the defining traits of MiniMax's product development philosophy.

In October 2022, one month before ChatGPT was launched, MiniMax released the predecessor of Talkie: a direct-conversational AI agent product named Glow. In June 2023, following Glow's trial operation in the domestic market, Talkie became MiniMax's first internal product to "go global on its very first day".

At that point in time, going global had not yet become a consensus across the AI industry.

To race against time and launch the product, MiniMax did not hesitate to release it first and fix flaws later, even when its AI technical capabilities still had imperfections.

Around October 2022, MiniMax was still conducting internal testing on its self-developed large language model with 30 billion parameters. A front-end engineer at MiniMax recalled to *Intelligent Emergence* that the model had numerous bugs at the time, which required front-end engineering work to compensate for.

He gave an example: since users liked to use "..." to express speechlessness, the model mistakenly assumed that all punctuation marks should be repeated three times. "Back then, we could only write more instructions. The model itself wasn't that smart," he said.

But MiniMax's brilliance lies in the fact that, unlike many products on the market that exist merely to show off their "AI muscles" — a phenomenon that remains widespread even today — the core of Talkie is not AI itself, but the effort invested in product positioning and user experience.

"Talkie is not really an AI application; it's a companion game," a senior AI product manager at a major tech company commented to *Intelligent Emergence*. "From the results, Talkie is not an application built to showcase AI features; the AI capabilities only serve to enhance its companion attributes."

The companion attribute, to a large extent, ensures that Talkie boasts exceptional user stickiness.

"Talkie's product logic is the same as 'raising a virtual pet' (companion dolls) and *Travel Frog*," a Talkie product manager told *Intelligent Emergence*. "The role of AI is to make the product — as users spend more time using it — create characters whose feedback feels increasingly 'understanding of you'."

He gave an example: once, he mentioned to his self-created AI character that he wanted to travel to Iceland before the end of the year. Recently, when he talked to the AI character about his annual leave, the AI asked: Did you finally fulfill your dream of going to Iceland?

The aforementioned AI product manager also shared a set of data: Talkie users who use the app continuously for a week have a payment rate and retention rate of around 80%.

A product that seizes a blank market with first-mover advantage can capture the attention of the first wave of users, which also means investing more effort into refining product features.

In the eyes of many industry insiders, Talkie has put more work into engineering and UI design than many other AI companion products. This is why, upon its release, Talkie remained a mature product that could attract users despite its technical flaws.

The aforementioned AI product manager at the major tech company has downloaded almost all well-known AI role-playing applications on the market, such as Character.AI and Replika, on his phone. He commented to *Intelligent Emergence* that most companies can reverse-engineer the engineering and functional design of Character.AI, but Talkie is extremely hard to replicate:

"We spent a lot of effort researching why Talkie placed a paywall here, or added an interaction there, only to find that its engineering is incredibly meticulous, with every part interlocking."

There are many products that have gained first-mover advantage, but to sustain that advantage, rapid iteration based on market and user feedback is essential.

"Every design choice in Talkie is iterated quickly based on user feedback," a Talkie employee told *Intelligent Emergence*. He recalled that Talkie had the largest number of design requirements across MiniMax. At peak times, the product team could propose more than 20 requirements, which around 5 design employees had to fulfill within a week. "The intensity was higher than at many startups I worked at before," he said.

Similarly, Talkie and its domestic version "Stellar" are the products to which MiniMax allocates the most manpower. The aforementioned employee revealed that Talkie and Stellar have nearly 300 employees in total, while the team for MiniMax's productivity tool "Conch AI" has no more than 100 people.

Almost every Talkie employee can cite examples of Talkie's meticulous design to *Intelligent Emergence*:

The UI deliberately uses gold and black to create a more luxurious feel; star elements are subtly placed in various places, such as in the eyes of some characters and at every payment checkpoint; the dialog boxes with characters automatically add parentheses, because users like to add "()" in conversations to express their unspoken asides...

Talkie AI character chat interface.

Iterating the product closely following user feedback means that even if other products on the market copy Talkie's entire design, they can hardly replicate its success.

"What I admire most about Talkie is that it stays focused, and all its features are built to serve its target users. Once the user base changes, Talkie's strategy stops working," an AI product manager at an AI unicorn commented. "Many current consumer-facing AI applications are too greedy, trying to implement every possible feature right from the start, which ultimately leads to low user activity and poor stickiness."

A Success That Defies Replication

But for an AI unicorn with a valuation of over 200 billion yuan that relies almost entirely on Talkie as its sole "cash cow", there is an underlying risk of total collapse if that single product fails.

The first wake-up call came in March 2023, when MiniMax hit a major hurdle: Glow, which had been launched only six months earlier, was removed from app stores.

Although the official did not disclose the specific reason, and the AI filing regulations had not yet been introduced at that time, many industry insiders and MiniMax employees believed that the incident was related to the chaotic user ecosystem on Glow.

Over-accommodating users allowed MiniMax to taste the sweet success of a consumer-facing product, but it also laid hidden dangers for its domestic development.

"Desire is human nature. Adult-oriented content is currently the most successful business model in AI," an investor once commented to *Intelligent Emergence*. Looking at the AI products that have seen significant growth overseas, such as Replika and Civitai — which are listed in the "Top 10 Fastest-Growing AI Applications of 2023 (Source: Writerbuddy)" — all have varying degrees of tolerance for adult and suggestive content.

To maintain this growth momentum, Talkie could only move to overseas markets with looser regulatory environments. However, none of MiniMax's domestic products have ever replicated Talkie's glory.

Take its domestic version "Stellar", launched in September 2023, for example. It basically replicates all gameplay mechanics of Talkie, and even has a more refined UI. Back then, to adapt to the European and American markets with the highest payment power, Talkie's design was not overly polished. A former Talkie employee told *Intelligent Emergence* that mobile phone performance overseas is generally lower than in China; too many engineering details would cause the app to lag, which would undermine the user experience for European and American users.

But Stellar has not gained the level of popularity that MiniMax internally expected in the domestic market.

"To pass regulatory filings, Stellar invested a lot of effort in setting up firewalls for sensitive words," the aforementioned MiniMax product manager mentioned. "But this seriously hurts the user experience. Sometimes users are going through a tough time and want to vent with a few curse words, only to have the AI lecture them to stay positive."

Statistically, female users are more willing to pay for companion products than male users. A former Talkie employee told *Intelligent Emergence* that the male-to-female user ratio is 2:8 for Character.AI, 4:6 for Talkie, but 6:4 for Stellar. "There is a clear gap in users' willingness to pay," he said.

Meanwhile, Talkie in overseas markets cannot rest on its laurels and take its market position for granted.

Anxiety is the emotion that many MiniMax employees have sensed from YAN Junjie since the beginning of 2024.

"The risk of relying on a single revenue source is too high. Right now, Talkie is effectively the heart of MiniMax," a Conch AI employee told *Intelligent Emergence*. She mentioned that the US ban on TikTok deeply unsettled the company's senior management, who feared that Talkie could become too high-profile in the US and suffer the same fate.

Such concerns are not unfounded.

On October 17, 2024, a Bloomberg report pointed out that as Chinese AI applications become more popular in the US, they are increasingly likely to face the same geopolitical issues as TikTok.

This year, MiniMax has spent significant effort moderating content on Talkie. Multiple aforementioned MiniMax employees all told *Intelligent Emergence* that through certain engineering measures, Talkie has gradually become unresponsive to suggestive conversations.

"In the long run, products that rely on suggestive content cannot retain users. These users only use the app for that kind of content, and will easily switch to another platform, so their stickiness is very low," a Talkie product manager mentioned to *Intelligent Emergence*.

Building another Talkie internally has also been put on MiniMax's agenda.

The aforementioned Conch AI employee told *Intelligent Emergence* that in mid-2024, the company once required Conch AI — a productivity tool — to add companion features that allowed users to create their own AI characters. However, this decision has not yet produced the expected results.

Multiple insiders revealed that the product's wavering positioning between a productivity tool and a companion tool led to the collective departure of Conch AI's product operation team; with strong competitors such as Kimi and Doubao in the market, Conch AI's user retention is currently far from ideal.

At the same time, MiniMax is not putting all its eggs in one basket, and has started to explore new markets beyond Europe and America for Talkie.

An insider disclosed to *Intelligent Emergence* that in March and April 2024, YAN Junjie mentioned at the weekly all-hands meeting every Friday that Talkie should start deploying in the Southeast Asian market, and that product design and features need to be adapted to local market needs.

"Talkie's growth in English-speaking markets has entered a stable phase, and there are still many opportunities in Asian markets with non-mainstream languages," he said. He also mentioned that Southeast Asia's advantages lie in lower channel and marketing costs, as well as a more tolerant attitude towards nurturing-type products.

Continuing to write its growth story in the AI era is the key for MiniMax to secure a solid position in the fiercely competitive AI industry.

Feel free to reach out for discussions!