Since the beginning of the year, the Chinese artificial intelligence firm DeepSeek has emerged as a focal point in the global technology and venture capital landscapesOn January 20, the company launched its DeepSeek-R1 inference model, which surprisingly achieved performance levels comparable to OpenAI's renowned models at a fraction of the cost.

This groundbreaking advancement has not only sent shockwaves throughout the industry but has also amplified the pressure on other leading model manufacturers, particularly those in the so-called "six small tigers" category, who have relied heavily on venture capital to propel their way into the final stages of the competition.

Wang Sheng, a partner at Inno Angel Fund, remarked to reporters, "In a sense, the emergence of DeepSeek has put the brakes on the frenzied race of large model startups we've seen over the past two years." He expressed that the initial enthusiasm for developing ultra-large-scale computing power-driven models has dimmed, revealing that the returns on such massive investments have not been as substantial as originally anticipatedDeepSeek's unique approach has encouraged industry players to pause and reconsider their technological and business strategies.

Looking back, we see a progression from the "hundred model battle" in 2023 to the ascendance of the "six small tigers" in 2024, culminating in DeepSeek's sudden rise in 2025. As AI continues to empower countless industries, the competition among model providers is intensifying, leading to an accelerated elimination processIn the wake of DeepSeek's influence, the commercialization and financing landscape for the "six small tigers" has become increasingly challenging, prompting them to actively adjust their strategies in search of breakthroughs.

One firm adapting to the changing landscape is Yuezi Anmian, which has opted to reduce its marketing expenses while prioritizing investment in application developmentFrom a consumer standpoint, the excitement surrounding DeepSeek has reshaped public perception of the AI industry, effectively educating the market

Advertisements

Prior to DeepSeek's debut, Yuezi Anmian's Kimi was the hottest AI application from a startup in China, as showcased by QuestMobile data, which highlighted that by December 2024, the top three native AI apps were Doubao, Kimi, and Baidu's Wenxiao Yan, with monthly active user numbers of 75.23 million, 21.01 million, and 12.24 million, respectively.

DeepSeek swiftly altered these dynamics, with QuestMobile reporting that on January 28, it exceeded Doubao's daily active user count, making it the top domestic model in this regardBy February 1, its daily active user count surpassed 30 million, achieving this milestone faster than any other application in history.

"Yuezi Anmian previously invested heavily in marketing and customer acquisitionHowever, DeepSeek managed to achieve an impressive user surge with minimal marketing spendIn terms of consumer application market performance, Yuezi Anmian has been the most impacted among the six small tigers," an investor closely observing the AI sector conveyed.

Nevertheless, insiders close to Yuezi Anmian revealed that Kimi has also experienced significant organic user growth recentlyThe company’s co-founder Zhang Yutong shared on social media that Kimi witnessed a marked increase in user numbers starting from the first day of the new lunar year.

Commercially, Yuezi Anmian had previously attempted to explore monetization on the consumer side through a tipping model, while Wenxin Yiyan offered a free basic version with a charge for the professional versionFollowing the emergence of DeepSeek, Wenxin Yiyan announced on February 13 that it would become completely free starting April 1, 2025. This could suggest that achieving profitability through a fee-based model for consumer AI applications has become more daunting in the current market environment.

In response to the market disruption wrought by DeepSeek, Yuezi Anmian has proactively revised its strategyReports on February 18 indicated that the company planned to significantly cut its product marketing budget, halting various advertising campaigns on Android platforms and collaborations with third-party advertising platforms

Advertisements

Additionally, following a team review, it was determined to focus on maintaining state-of-the-art foundational modelsA re-training of their core model is anticipated as the organization seizes opportunities in application development.

"At this stage, DeepSeek has positioned itself at the industry's forefrontFor Yuezi Anmian, continuing to invest in the market at the previously maintained pace would yield limited benefitsReducing expenditure on marketing, taking a step back to preserve financial strength, and seeking new breakthroughs is a rational choiceI believe Yuezi Anmian is strategizing something significant, aiming to deliver high-performance, low-cost applications," Zhang Yu, founding partner at Qingzhi Capital, analyzed.

Interestingly, many coincidences exist between DeepSeek and Yuezi Anmian: their founders, Liang Wenfeng and Yang Zhiling, both hail from GuangdongYuezi Anmian's headquarters is located in JD Technology Tower, while DeepSeek's development team operates from an office in the Rongke Information Center, situated less than two kilometers apartFrom Yuezi Anmian's conference room, one can even glimpse the silhouette of the Rongke building.

In terms of product launches and academic publications, the firms appear to intentionally synchronize their efforts — on January 20, both DeepSeek and Yuezi Anmian released their next-generation models, DeepSeek R1 and Kimi K1.5, respectivelyBy February 18, they almost simultaneously published research papers focusing on challenges to the core attention mechanism of the Transformer architecture, enhancing its efficiency in processing longer contexts.

"Such coincidences may not be entirely accidental; they potentially indicate that both companies are closely monitoring each other's movements amidst fierce market competitionThe potential for consumer applications in large models remains substantial, and Yuezi Anmian is not without an opportunity to catch up, but it must adopt a long-term perspective on market trends," an investor opined.

Meanwhile, among the "small tigers," differentiated strategies are being pursued to navigate industry applications

Advertisements

In the B2B market, DeepSeek’s characteristics of low cost, high performance, and open-source nature render it similarly competitiveIn fields like government and finance, the focus on data privacy and the need for local deployments are pressingDeepSeek's open-source model dramatically lowers the costs associated with local deployment of top-tier models.

Currently, multiple local government systems have integrated DeepSeek, welcoming partnerships with numerous financial institutions and state-owned enterprisesFor the smaller firms primarily focused on B2B and B2G business, the pressure from these developments is palpable.

To illustrate, Zhipu AI, recognized for its commercialization capabilities, disclosed at the end of last year that its commercial revenue had achieved over 100% growth compared to 2023 due to its newly launched multi-modal platform, bigmodel.cn, which now attracts 700,000 enterprise and developer usersZhipu AI's offerings span intelligent cars, manufacturing, consumer goods, finance, public services, and healthcare, with their Qingtong APP boasting over 25 million users and an annual recurring revenue exceeding ten million.

"The advent of DeepSeek has led many state-owned enterprises to opt for private model deployments using DeepSeek technology, which will undoubtedly impact Zhipu AI's B2B operationsHowever, Zhipu AI's strengths lie in its robust multi-modal capabilities and early investments in code generation, augmented by its low-code platform, further expanding its competitive edge." Zhang Yu articulated.

Reports indicate that Zhipu continues to develop along multi-modal lines, planning to release several new products in 2024. In 2025, the keyword "Agent" will center Zhipu’s development — they recently announced a collaboration with Samsung to integrate their Agentic GLM into the Galaxy S25 series, enhancing AI-driven real-time voice and video calls along with visual understanding and system functionalities.

Furthermore, numerous companies are integrating multiple large models

For instance, Hesai publicly announced its integration with DeepSeek, Minimax, Tongyi Qianwen, Yuezi Anmian, and DoubaoHesai’s founder and CEO Ma Chuanquan stated that as a SaaS tech firm, they are eager to explore diverse connections across various models.

"On one hand, disparate client preferences drive model selection, necessitating matches with existing deploymentsOn the other hand, different models showcase unique features, allowing application across various business scenarios," Ma commented.

Other small tigers are also carving their own paths into industry-specific applications, helping them sidestep the impasse brought by DeepSeek’s competitive sharpnessFor instance, Lingyi Technology's CEO, Kai-Fu Lee, stated in January that the company had established an "Industry Large Model Joint Laboratory" in collaboration with Alibaba CloudSubsequently, Lingyi Technology decided to abandon super large model training, instead focusing on developing medium-sized models that are faster and more cost-effective for generating profitable applications.

On February 14, Lingyi announced the official establishment of an industrial large model base in cooperation with Suzhou High-tech Zone, exploring model deployments across different industrial domains in collaboration with companies like Zhongxi Soft Group and Hypermedia Group.

Bai Chuan Intelligent has begun concentrating on the medical sectorThe founder and CEO, Wang Xiaochuan, publicly declared that healthcare represents the "crown jewel" of large modelsBai Chuan aspires to become a leading large-model firm while achieving breakthroughs in the medical sector to develop super applications.

In a contrasting dynamic, Leap Star, the smallest finanicng entity among the six small tigers, has made notable strides in vertical industriesWith prior efforts in finance, content creation, and smart devices, Leap Star recently announced a collaboration with Geely Auto Group to jointly open-source their Step series multi-modal models to global developers.

“Focusing on vertical industries and forming in-depth partnerships with leading firms is an excellent development strategy

Vertical industry data can create significant barriers, allowing large model companies to establish differentiated advantagesIn essence, large model applications across various sectors are just beginning their journey; collaboration in exploring the market far exceeds the competition narrative,” an investor suggested.

Ji Kaiming from Yunxiu Capital emphasized that the competition intensifies for the six small tigers, as financing has solidified over the past year, broadening their cash reservesIn response to DeepSeek's impact, these smaller firms have pivoted their focus toward differentiation while exploring model applications as a cohesive unit, creating barriers in data, industry, customer relations, and product operations that will afford them better development opportunities.

Moreover, Minimax has shifted its focus towards international markets, carving out a niche in AI-driven emotional companionshipMinimax offers apps like Shell AI, Xingye, and Talkie (overseas), targeting C-end users, and has also launched an open platform drawing in over 30,000 B2B clients and developers.

In stark contrast, the financing situation for the "six small tigers" and DeepSeek is telling of a sector undergoing survival of the fittest.

In the early stages of AI model expansion, investments were speculative, with firms backed primarily by academic credentials and technical expertise betting on their future potentialAfter two to three years of fierce competition, the AI model landscape crystallized around the "six small tigers." However, the industry was surprised by DeepSeek's thrilling product capabilities, prompting investors to reassess their strategies moving forward.

The challenge for the "six small tigers" now lies in proving their capacity for revenue generation and user growth, which will be essential in attracting further investment. "Raising funds will become more challenging, potentially leading to lower valuation expectations as the competition heats up," Wang Sheng stated.

Conversely, DeepSeek’s position in the funding arena is burgeoning

Advertisements

Advertisements