In recent times, we have witnessed a significant shift in the pricing strategy of large model vendors, marking the end of the era where costs were calculated on a per-unit basisThe new trend leans towards substantial reductions in prices, with reports of discounts as steep as 97%. This stunning development has left many in the tech world intrigued and pondering the implications of such pricing strategies.

On May 25, a report by Huafu Securities indicated that the price cuts were initiated by Deepseek-V2, a model under the renowned private equity giant, Huanfang QuantitativeIn response, major players in the industry including ByteDance, Tencent, Alibaba, and iFLYTEK have followed suitTo date, seven companies have collectively reduced the prices of 21 different large models, sparking what many are calling a pricing revolution.

When asked about the driving forces behind this trend, Zhang Yi, CEO of iiMedia Consulting, noted in an interview: “It’s not just domestic vendors that are lowering prices

Companies like OpenAI abroad are also slashing costsThe logic behind this is quite straightforward—price reductions are often a response to competitive market pressures.”

The catalyst for this collective price-cutting movement stems from the price adjustments introduced by Deepseek-V2 on May 6. The model set the market abuzz by offering input and output costs of just 1 and 2 yuan per million tokens, respectivelyWhile Deepseek-V2 is designed to compete with models like GPT-4 in performance, its pricing stands at a mere 1% of that of its competitors, making it a truly cost-effective alternativeFollowing this move, other models such as Zhipu and Doubao have also joined the ranks of those offering lower prices.

It is particularly notable that the reductions have been most pronounced for lightweight and secondary models, while primary models at the GPT-4 level also experienced some degree of cost decrease

• OPEC+ Delay: What Lies Ahead for Oil Prices?
• Gold Weakens and Declines
• Gold & Silver: Volatile Ahead of Non-Farm Payrolls
• Hedging Options as Bitcoin Breaks $100K Barrier
• Coal Power's Low-Carbon Shift: External Aid & Internal Upgrades

Analysts from Huafu Securities predict that with further price decreases, B-end clients, especially existing customers of cloud service providers, will expedite their adoption of large model products, paving the way for accelerated commercialization.

During a recent interview, Shen Dou, Executive Vice President of Baidu Group, explained the rationale behind offering their Wenxin series of large models for freeHe emphasized that the objective was simple: “We hope to steer away from daily comparisons of pricesWith the reduced costs, companies can be more adventurous with innovation and quickly validate their ideas.” Shen attributes the possibility of lowering prices to Baidu’s advanced technological capabilities, particularly their efficient AI heterogeneous computing power management platform and multi-core solutions.

Industry insiders note that the current landscape for large models reveals a tremendous commercial opportunity reminiscent of the search engine boom over twenty years ago

Both the current capabilities of large models and the problems they aim to address bear striking similarities to those solved by early search engines, but now large models offer more robust analytical abilities, catering closely to user demands.

From this perspective, these products open the door for platforms to potentially attract billions of usersAs Zhang Yi points out, if a product has the potential for massive user acquisition, implementing membership fees does not align with its business logicMuch like how WeChat chose not to charge its extensive user base, citing that the value generated through services such as gaming and advertising far surpasses any membership income.

Regarding the ongoing price drops among large model companies, Zhang Yi explained the rationale: “Charging users can lead to them opting for competitorsThus, in a highly competitive market, acquiring a larger user base becomes the primary development goal.” He mentioned the inevitability of lower prices and free access as businesses scramble for survival before potential adversaries emerge.

Additionally, Fu Sheng, Chairman and CEO of Cheetah Mobile, remarked that the current drop in prices serves as a wake-up call for startup companies in the large model sector; they must rethink their business strategies

He highlighted that major cloud service companies withdrawing their prices significantly pursue more cloud clients, while smaller firms need new business models to succeed.

The effects of these price drops are profoundHuafu Securities expects that as large model prices continue to decline, B-end clients—especially existing cloud customers—will be primed for accelerated product adoption, fostering speedier commercial realization.

Liu Weiguang, Senior Vice President and President of the Public Cloud Division at Alibaba Cloud, articulated a goal behind the sharp drop in inference costs: to drive an explosion in AI applications and hasten their commercial deploymentNew data from Alibaba Cloud indicates that the Tongyi large model has already been utilized by over 90,000 companies through its services and by over 2.2 million entities via DingTalk, effectively spanning various sectors including education, healthcare, and gaming.

The lowered pricing strategy by cloud vendors holds significant value for smaller enterprises, which often operate with limited budgets

These businesses can now experiment and innovate more freely with large model applicationsSome analysts speculate that as cloud providers reduce or remove fees for premium models, smaller companies are likely to transition from open-source models to high-quality closed-source ones due to cost efficiency.

One ecosystem sales leader from a cloud vendor claimed that massive numbers of independent small application development teams will emergeTeams comprising three to five members can now swiftly develop engaging applications, blending AI and SaaS platforms in innovative waysCertain AI applications are expected to replace conventional SaaS products, while others may integrate as components within SaaS applications, reminiscent of the mini-programs in popular apps like WeChat and Alipay.

Ultimately, the reduction of overhead costs paves the way for explosive growth in application-based companies

Historical trends in the mobile internet sector highlight this—only after significant drops in telecom and cloud service expenses did giants like WeChat and Douyin emerge.

Within the large model arena, a similar evolution could be imminentToday, while Chinese firms lag behind leading U.Sentities like OpenAI in foundational model capabilities, numerous investors believe that opportunities for China to leapfrog competitors lie within applications of large models.

The burgeoning ecosystem of AI applications is also evident in Silicon Valley, where strong base models empower small startup teams to quickly generate revenue and secure funding.

Experts within the industry foresee that the ongoing price cuts could trigger a shakeup, causing weaker firms with low-cost structures and limited financing to exit the marketThis transformation could further encourage technological innovation, as lower costs will enable businesses to meet complex needs effectively while validating the applicability of large models.

From the users’ viewpoint, decreased usage costs might catalyze a swell in large model users, particularly in the consumer market, alongside a notable rise in computational demand