The recent divergence in market valuation between MiniMax and its rival Z.AI has raised serious questions, with the latter now commanding a market value roughly 3.9 times that of MiniMax. Despite both companies targeting similar customer segments and achieving comparable revenue figures, investors appear to be voting with their feet against MiniMax, and the underlying causes are complex.
According to its interim results for fiscal 2026, MiniMax reported revenue of approximately $117 million in the first half of the year, a year-over-year surge of 283.1% and already 1.5 times its total revenue for all of last year. Its gross profit also improved dramatically, climbing 464.8% to $20.81 million, thanks to enhanced infrastructure and efficiency gains. However, the company still recorded a net loss of $358 million for the period, a narrowing from the $402.2 million loss in the same period last year, while its adjusted net loss rate improved from roughly 455.9% to 251.4%.
Interesting to note, MiniMax's market value once temporarily surpassed that of Z.AI after the latter went public as the first global large model stock. Yet by the close of trading on September 15, 2026, Z.AI shares stood at HK$680, with a market cap of about HK$316.6 billion, while MiniMax shares were at HK$234.6, valuing the company at just around HK$81.9 billion.
Why is MiniMax underperforming, despite its global user base for consumer products?
MiniMax has leveraged its native applications, including Talkie/Starsea and Hailuo AI, to build a substantial international user base, with over 70% of its revenue coming from outside China. For full-year 2025, it generated $79 million in revenue, with AI-native product revenue hitting $53.08 million, up 143%. By the end of 2025, it had served over 236 million users across more than 200 countries and regions, along with 214,000 enterprise clients and developers worldwide.
Initially, this global narrative fueled MiniMax's strong debut on the stock market, with its public offering oversubscribed by 1,837 times and a first-day surge of 109%. In contrast, Z.AI only managed a 13.17% gain. However, the story has since soured. While daily token call volume in China has exploded from about 100 billion in early 2024 to over 500 trillion by mid-2026, MiniMax's C-side revenue growth has not kept pace. Its AI-native product revenue grew by 100.9% to $42.6 million in the first half of 2026, but this reflects a user base that may be showing signs of stagnation; for instance, the MAU for Starsea actually dipped by 3.21% month-over-month to 3.71 million in June.
Worse, the gross margin for its C-side AI-native products is painfully thin at just 4.7%, meaning for every dollar of revenue, the company earns less than five cents in gross profit. This raises doubts about whether MiniMax can cover the high inference and marketing costs associated with its consumer apps.
Is MiniMax falling behind in model capability?
Industry insiders suggest that pricing power in the AI sector now hinges on technological leadership. This is an area where MiniMax appears to be conspicuously lagging. According to the official profile from Z.AI, its GLM-5.3 model has achieved a score of 60 on the Artificial Analysis Intelligence Index (AAII), placing it in the global frontier model range, alongside top-tier closed-source models like Claude Fable 5 and GPT-5.6 Sol, and tied with Kimi K3 for the best open-source model. MiniMax-M3, however, sits outside the top ten, scoring just 45 points, a significant gap.
This capability advantage has translated directly into commercial success for Z.AI. In Q1 2026, it raised API pricing by 83%, yet still saw volumes surge. Its API gross margin has also turned positive, reaching 24.6%, demonstrating that its unit economics have crossed the break-even point, driven by scale, system optimization, and lower inference costs. Per the company's data, its per-token inference cost has fallen about 80% since the start of the year, and revenue per yuan of computing investment has risen 14-fold year-over-year.
In the rapidly growing Token revenue ranking, Z.AI has clearly entered the first tier due to its AI Coding strategy, alongside ByteDance and Alibaba, while MiniMax is relegated to the second tier, competing with DeepSeek, Kimi, and Keling. The second tier also shows contrasting strategies: DeepSeek relies on API price increases, Kimi on its strong single model, while MiniMax and Keling depend on video models to expand overseas.
Does MiniMax B-side growth rely on "low price for volume"?
To compensate for its lagging C-side growth, MiniMax has aggressively pivoted to the B-side. Its B-side revenue skyrocketed 703.1% year-over-year to $73.9 million in the first half, now making up 63.4% of total revenue, a significant shift from just 30.3% in the same period last year. Its enterprise customer and developer count has also exploded to over 2 million, a tenfold increase. However, the quality of this growth is dubious. Public records show MiniMax engaged in a "50% price cut + billing model change" strategy. On June 1st, upon the release of M3, it switched to token-based billing and canceled its cheap $29/month Starter plan, raising the entry price to $49. This triggered a user backlash and complaints, especially on V2EX, forcing the company to issue an apology and provide compensation. It then permanently slashed prices by 50% on June 15th.
Analysts note that MiniMax's language model M3 has failed to meet expectations, consuming significant computing resources without delivering corresponding returns, which negatively impacts its overall gross margin. While its video model H3 has opened new horizons - with a 5-second generation time beating the playback speed and pushing ARR up from 3 billion in May to 4.8 billion by August - the broader model strategy appears to be under heavy pressure on both the C and B sides. This suggests that behind the impressive revenue numbers, MiniMax's unit economics may still be at an inflection point where it is trading profitability for scale, a fact that might explain the market's current decisive preference for its rival.