The rapid ascent of the DeepSeek-R1 inference model has taken the artificial intelligence landscape by storm, stirring conversations about its potential and performance just a month after its launchThis innovative model, with its API costs only a fraction—merely 3%—of what it takes to run the OpenAI o1 model, manages to deliver performance that rivals the latterSuch a combination of low cost and high efficiency marks a significant milestone in the quest for accessible and effective AI solutions.

One of the most notable aspects of DeepSeek is its commitment to open sourcing the R1 model entirelyThis move has been characterized as a game changer by observers within the industry, including Zhou Zhengang, the vice president of IDC ChinaZhou claims that the open-source nature of DeepSeek significantly lowers the barriers for widespread AI adoption, alluding to an impending surge in artificial intelligence applications across various sectors.

The signs of this trend are starting to showIn a matter of weeks, players throughout the AI ecosystem, from chip manufacturers and cloud service providers to downstream applications, have been rushing to integrate DeepSeek into their offeringsThe consumer market has also displayed a burgeoning interest in the model; data from QuestMobile reveals that as of February 1, DeepSeek reached an impressive daily active user count of over 30 million, making it the fastest application to achieve such a milestone.

With the rise of DeepSeek comes a pivotal question for other major AI model developers: how will they respond in the face of this formidable competition? The “six small tigers” of the Chinese AI scene—companies like Zero One Everything, Baichuan Intelligent, StarStep, Zhipu AI, The Dark Side of the Moon, and MiniMax—have been particularly affected by these developments.

In the lead-up to the release of DeepSeek-R1, the general mood in the Chinese AI market had been one of concern and skepticism

Advertisements

Manufacturers found themselves grappling with supply restrictions of Nvidia chips, which severely limited their computational capabilitiesEven for those who were able to acquire sufficient hardware, the steep costs associated with high-performance computing began to weigh heavily on their operationsOne major player in the industry famously expressed during a company-wide meeting that they wished GPT-5's release would be postponed, essentially out of budgetary constraints—a sentiment that, while exaggerated, reflected the anxiety permeating the sector at that time.

As early as 2023, the much-anticipated "battle of the hundred models" that dominated discussions turned out to be short-livedBy 2024, only a handful of companies remained in the game, still laboring over pre-training modelsAmong the notable players, Zero One Everything stood out as the first to pivot away from pursuing increasingly large-scale modelsIn early 2025, they announced a partnership with Alibaba Cloud to establish a joint lab focused on industrial models, with CEO Kai-Fu Lee noting that only the larger corporations could sustain the development of colossal models moving forward.

Another example of the industry shift can be seen in Baichuan Intelligent, which has taken a focused approach towards healthcare-related AI solutionsOn February 13, they launched the "AI Pediatrician" based on their Baichuan-M1 model, following a month of internal testing in collaboration with Beijing Children's Hospital.

In a somewhat contrasting approach, The Dark Side of the Moon's efforts to release their Kimi K1.5 model coincided with DeepSeek-R1's launch dateHowever, their announcement garnered little attention in comparisonReports have indicated that this company is now reevaluating its strategic direction in light of DeepSeek's swift riseReflecting on a year of missed opportunities, a researcher from The Dark Side voiced their regret over prioritizing long-text capabilities due to cost concerns, hinting at the inherent risks faced by start-ups caught in a competitive landscape.

The remaining players, known informally as the "three little tigers," are adapting differently

Advertisements

Zhipu AI is focusing on increasing its enterprise interactions with their models, while MiniMax and StarStep continue to release their latest multimodal modelsNotably, both companies have chosen to integrate DeepSeek into their offerings, highlighting the impact of its rapid growth.

As DeepSeek's influence expands, concerns about the future pathways for the "six small tigers" loom largeKai-Fu Lee has remarked that when pre-trained results fail to surpass open-source models, there is little justification for continuing the pursuit of pre-training—particularly for start-ups, where the commercial viability remains crucial.

With advancements in model performance and decreasing inference costs, many industry analysts view 2025 as a critical turning point for commercialization within the realm of large AI modelsAs the landscape shifts, the strategic focus of these players will likely pivot towards practical applications and optimizing their business models, making this year a defining moment for their prospects.

Bigger companies, including ByteDance, Alibaba, and Tencent, also find themselves in the crosshairs of DeepSeek's rapid advancementA ByteDance employee recounted how their team was abruptly called back from a team-building event to strategize in light of the DeepSeek-R1 launch.

Another employee from a major firm revealed that their management team views the general pursuit of large models as narrowing, positing that future opportunities would favor larger corporations while the smaller players would merely be testing grounds for the innovations of these giantsYet, DeepSeek's emergence suggests there may be more avenues for competing players outside of this narrative.

DeepSeek distinguishes itself from the "six small tigers" by showcasing an approach that merges startup ingenuity with the backing of significant industry resources akin to larger firmsWith no funding rounds since its inception in 2023, its transition into the market leverages robust support from its parent company, an established quantitative powerhouse named Huansquare—granting DeepSeek ample financial resources to pursue ambitious General AI (AGI) objectives without the immediate pressure of generating predictable returns.

The successes of DeepSeek may thus fuel rivalry among established players, with questions arising about the potential breakthroughs attainable by companies with ample resources

Advertisements

Advertisements

Advertisements