MiniMax Group Inc. engages in research and development of an artificial intelligence (AI) foundation model. The company's platform offers a wide range of modalities, including language, video, speech, and music. It also provides AI-based enterprise services; AI-native products, including enterprise-facing open platform; and consumer products, such as MiniMax Agent, Hailuo AI, Talkie and Xingye. The company was founded in 2022 and is headquartered in Shanghai, the People's Republic of China.
Country
Sector
Themes
Also in
Price· split & dividend adjusted
No price history for this asset yet.
News & notes moving0100.HK
Artificial Intelligence▲impact 4
Z.ai shares surge 8% after releasing AI model running on Chinese chips
Chinese artificial intelligence company Z.ai released a new model Wednesday that it claims operates entirely on homegrown semiconductors, sending its Hong Kong-listed shares up more than 8% in Thursday trading. The low-cost model, called GLM-5.3-Flash, ranks 10th on the Artificial Analysis Intelligence Index, ahead of DeepSeek V4 Pro Max. Z.ai said it used 100,000 China-made chips to handle all online requests for the model, which was released on Aug. 20 under the code name "Ox Alpha" and ranked first by usage on the global OpenRouter platform in the last week. CNBC could not independently verify the chip claims, and Z.ai declined to name the chip suppliers. The release comes amid U.S. restrictions on advanced chip sales to China, prompting domestic alternatives from Huawei and others. Separately, rival MiniMax's shares rose about 3% in Hong Kong after reporting a 283% revenue surge in the first half, though its adjusted net loss more than doubled to $293 million. Z.ai is scheduled to report its first-half results on Monday.
The General Office of the Shanghai Municipal People's Government issued the 15th Five-Year Plan for the Development of Strategic Emerging Industries in Shanghai, proposing to further comprehensively elevate the capability level of the integrated circuit industry, accelerate the research, development, and industrialization of high-end chips, and focus on developing high-end chips, AI-empowered design, advanced packaging, novel devices, and new-architecture chips. Meanwhile, CXMT announced that the over-allotment option for its initial public offering has been fully exercised, corresponding to an additional issuance of 1.003 billion shares, bringing total share capital to 67.884 billion shares. Hua Hong Semiconductor released its semi-annual report, with operating revenue of 9.574 billion yuan, up 19.41 percent year on year, and net profit of 399 million yuan, up 436.69 percent year on year. Lianxun Instruments saw net profit grow 903 percent year on year. In addition, MiniMax disclosed that its August annualized recurring revenue exceeded 800 million US dollars, and its token consumption in July reached 20 times the level at the start of the year.
DeepSeek officially releases flagship AI model V4 Pro, aiming for monetization with aggressive pricing
Chinese AI startup DeepSeek officially released its flagship AI model V4 Pro on the 13th. The price is set at several times that of the lower-tier V4 Flash, positioning it as a flagship model that leverages the high performance shown in benchmark tests to pursue monetization. According to independent benchmark firm Artificial Analysis, V4 Pro 0813 is priced at 1.32 dollars per million input tokens and 3.96 dollars per million output tokens, while V4 Flash, officially released last month, costs 0.14 dollars for input and 0.28 dollars for output, making Pro roughly 9 times more expensive for input and about 14 times more for output. V4 Flash had produced an unusual result in benchmark tests by outperforming the Pro preview version released in April, but the official V4 Pro has significantly improved performance over the preview. Artificial Analysis gave V4 Pro's reasoning mode a score of 53 on its Intelligence Index, which integrates nine evaluations including agent task execution, tool use, coding, scientific reasoning, and long-context handling, surpassing V4 Flash's score of 40. DeepSeek drew widespread attention in early 2025 when its R1 model showed performance on par with US players such as OpenAI despite low cost, but in China, emerging AI companies such as Moonshot AI, Zhipu AI, and MiniMax have since appeared one after another, eroding its advantage as a first mover.
AI Applications Surge Again, Chuanzhi Education Hits Seven Consecutive Daily Limit Ups, Brokerages Flag Two Industry Shifts as Opportunities
AI application concept stocks surged again on August 4, with the Zhipu AI segment leading the gains and Chuanzhi Education locking in its seventh consecutive daily limit up. By the midday break, UCloud had risen over 12 percent, Kaiwen Education recorded three boards in two days, and Pingzhi Information and IAT Automobile Technology both gained more than 9 percent. On the news front, US-listed AI application concept stocks Palantir and Snap both reported results that beat expectations, sending their shares up more than 14 percent in after-hours trading at one point. Guolian Minsheng Securities believes the education sector is poised for a three-dimensional resonance of policy, supply, and demand, with AI plus education as the main investment theme. Huaxin Securities judges that 2026 will be a key watershed for the AI industry, and media application scenarios are also expected to capture the dividends of the AI era. Guosheng Securities flagged two industry shifts as opportunities: first, major domestic and overseas players are collectively betting on Work Agent, and Work is likely to become the next deterministic direction for token consumption after Coding; second, in multimodal AI, ByteDance's Seedance 2.5 and MiniMax's Hailuo H3 have been released in quick succession, with the competitive focus extending to controllability, editability, and commercial licensing, and downstream enterprises in short dramas, advertising, and film and television are expected to be the first to benefit.
AI model distillation emerges as a new flashpoint in the US–China tech rivalry
Distillation, a technique that enables the downsizing and streamlining of AI models, has become a new flashpoint in the US–China battle for AI supremacy. Distillation uses the output of a large teacher model to train a smaller student model, and has long been regarded as a standard research method. But it is now drawing scrutiny over whether it can transfer AI capabilities without the consent of the developing company. The US government and leading American AI firms accuse Chinese competitors of using distillation to extract capabilities from US-made models. Anthropic claims that DeepSeek, Moonshot, and MiniMax carried out large-scale efforts to acquire capabilities from its Claude model. OpenAI has also stated that it detected attempts by Chinese-linked actors to use its models for distillation purposes. However, there have been no reported instances of Chinese companies accusing US firms of distilling their closed-source models.
MiniMax releases H3 video model with open-weight plan
Chinese AI firm MiniMax released its H3 video-generation model on Friday, capable of processing text, images, video, and audio to produce up to 15-second clips in 2K resolution with native stereo sound. The Shanghai-based company said it would release H3's model weights within days, extending the open-weight approach into video generation, and that generating 2K video would cost less than one-third of mainstream rival products. H3 is aimed at commercial applications including advertising, e-commerce, product design, and games, and was designed to work with several Chinese-made chips. MiniMax, one of China's well-funded 'AI tigers,' went public in Hong Kong in January and competes with ByteDance's Seedance 2.0 and Kuaishou's Kling 3.0 in the fast-growing video-model market.
Alibaba previews Qwen 3.8 Max AI model, shares jump in Hong Kong
Alibaba shares jumped as much as 5.4% in Hong Kong after the company previewed Qwen 3.8 Max, a new AI model it says ranks just behind Anthropic's Fable 5. The release came only days after Moonshot launched Kimi K3, adding to a burst of competition among Chinese developers building high-end coding and reasoning models. Alibaba has already made Qwen 3.8 Max available through platforms including Qoder, giving developers an early chance to test it. The market reaction also showed how quickly leadership can shift, with MiniMax falling 6% and rival Z.AI dropping as much as 15%. Goldman Sachs expects competition in coding models to intensify but still views Alibaba as its top cloud pick because of its heavier AI spending.
China's open-weight AI models dominate usage as Silicon Valley confronts premium pricing threat
Chinese open-weight AI models now occupy the top five spots on OpenRouter by weekly token usage, signaling a shift toward cheaper, customizable alternatives to premium American systems. The five models come from Tencent, Xiaomi, DeepSeek, MiniMax, and Z.ai, all allowing users to download, customize, and run them on their own infrastructure. This surge follows the release of Moonshot AI's Kimi K3, which rivals Anthropic's Fable and OpenAI's GPT-5.6 in key benchmarks, and comes as companies increasingly adopt lower-cost models for routine tasks like coding and summarization. Kong CEO Augusto Marietti noted that open-weight use has surged because flagship models are too expensive, while Mozilla CTO Raffi Krikorian compared using frontier AI for everyday work to driving a Ferrari to Whole Foods, adding that cheaper models can cost up to 50 times less. The trend threatens the premium pricing and blockbuster IPO plans of OpenAI and Anthropic, whose valuations depend on frontier AI remaining scarce and indispensable.
Palantir CTO Warns AI Distillation Threat Could Undermine U.S. AI Lead
Palantir Technologies CTO Shyam Sankar warned that Chinese developers may be building advanced AI models through unauthorized distillation attacks on U.S. technology, calling it a potential economic threat. Sankar said Chinese open-source systems appear to be products of distillation, where a lower-cost model is trained using outputs from another AI system, and argued that major American AI labs have a financial incentive to protect their intellectual property more aggressively. The warning follows Anthropic's accusation last month that Alibaba Group Holding conducted a large-scale distillation campaign through thousands of fraudulent accounts, adding to concerns that Chinese startups DeepSeek and MiniMax may have used similar methods to create rival chatbots at lower cost. Sankar also identified growing U.S. opposition to AI data centers as an even larger economic risk, citing a moratorium on new large data centers ordered Tuesday in New York by Governor Kathy Hochul, and said retreating from AI development could carry major long-term consequences for the United States. Separately, he did not blame France for replacing Palantir's data tools at its domestic intelligence agency with a local alternative, though he criticized what he described as unfair efforts to weaken Palantir's position outside the U.S.
MiniMax Slides 18% as JPMorgan Cuts Target Again Amid $2 Billion Raise
MiniMax Group Inc. shares fell as much as 18% on Monday after JPMorgan Chase & Co. cut its price target for the second time in less than a week, citing dilution concerns from the company's proposed $2 billion fundraising. JPMorgan lowered its target by another 20% following a 25% cut on Tuesday, while UBS Group AG halved its target on Sunday. The Shanghai-based AI developer plans to raise HK$9.5 billion from a share placement and HK$6.5 billion through zero-coupon convertible notes, which JPMorgan analysts estimate would represent a combined 17% of total shares upon full conversion. The selloff extends MiniMax's decline from its March peak to more than 80%, narrowing its gain since its January listing to approximately 40%, and follows a halving of the price of its M3 model one week after launch and the expiry of a six-month lock-up on IPO shares.
Minimax to raise approximately HK$15.957 billion; Zijin Mining and TCL Electronics forecast sharp profit growth
Minimax has announced plans to raise a net total of approximately HK$15.957 billion through a private placement and issuance of convertible bonds. Around 80% of the proceeds will be used to strengthen AI infrastructure and model research and development, with 10% allocated to the global expansion of its AI agent Harness. Zijin Mining Group has indicated that its net profit for the interim results ending June 2026 is expected to be approximately 39.1 billion yuan, up 68% year-on-year. TCL Electronics has forecast that its adjusted net profit for the same period will be between HK$1.48 billion and HK$1.65 billion, representing a year-on-year increase of 40% to 56%.
Tec-Do and MiniMax launch $100,000 MARS2 Multimodal Reasoning Competition at ECCV 2026
Tec-Do and MiniMax are sponsoring the MARS2 Multimodal Reasoning Competition with a total prize pool of $100,000 USD, held alongside the MARS2 Workshop at the European Conference on Computer Vision (ECCV 2026) in Malmö, Sweden. The competition features three tracks—Multimodal Advertisement Comprehension, Video Temporal Grounding, and Marketing Strategy Decoding and Conversion Analysis—designed to test complex, open-world reasoning and long-chain inference in enterprise-level environments. Keynote speakers include Paul Liang from the Massachusetts Institute of Technology, Mihaela van der Schaar from the University of Cambridge, and Louis-Philippe Morency from Carnegie Mellon University. The workshop also invites paper submissions on topics such as multimodal reasoning and agentic systems, with a submission deadline of August 1, 2026. The event will take place on September 8–9, 2026, and registration is hosted on the EvalAI Challenge Page.
Chinese AI stocks rally on demand optimism and Beijing policy support
Chinese artificial intelligence-related stocks surged after CSRC Chairman Wu Qing announced regulatory reforms to accelerate technological self-sufficiency. The measures, unveiled at the Lujiazui Forum, ease IPO standards for AI developers, support advanced sectors like quantum computing, and encourage Hong Kong tech companies to dual-list on the mainland. Hong Kong-listed LLM developers Zhipu and MiniMax jumped over 23%, while onshore chipmakers Cambricon Technology, MetaX, and Moore Threads also posted significant gains, driving the ChiNext Price Index to new highs. Analysts noted the rally highlights a stark divergence, with non-AI sectors remaining in a bear market.