Baidu Dazi, honored as the 'Exhibition Treasure' at WAIC, experiences a 20-fold surge in daily queries, capturing widespread attention.

On July 17, Baidu's versatile 'Baidu Dazi' was named one of the top ten 'Exhibition Treasures' at WAIC 2026, distinguishing itself as the sole versatile intelligent agent product among the winners.

Data released by Baidu reveals that Baidu Dazi has witnessed a 20-fold increase in daily queries and secured a top score of 93.3% in the PinchBench benchmark test. Notably, it showcased real-world business processes on-site, including Meituan errand services, cross-application task decomposition, and multi-terminal memory sharing, illustrating its evolution from 'chat-capable' to 'task-accomplishing'.

However, with fame comes scrutiny. The industry remains divided on whether this highly anticipated intelligent agent can surmount the various challenges encountered in practical applications.

Behind the Glory: Unstable Performance and Unverified Commercialization Data

Despite its 'Exhibition Treasure' status, 'Baidu Dazi' still exhibits signs of vulnerability in practical applications.

Immediate concerns revolve around product stability and user experience. On the morning of July 14, the mobile version of 'Baidu Dazi' experienced service delays, with some users reporting sluggish operations and slow responses, detracting from the overall user experience.

Baidu officially explained that the sudden spike in active mobile users, coupled with a significant rise in task assignments, caused temporary service capacity strain. The platform promptly initiated emergency capacity expansion and optimized database capacity and task processing pathways to restore normal operations swiftly.

For AI products reliant on real-time interaction, system stability and user experience are inextricably linked. Although this fluctuation was an isolated incident, similar issues could recur during future widespread commercial use, raising concerns about system stability.

Equally noteworthy is the uncertainty surrounding its commercialization.

To date, Baidu has not disclosed crucial operational data for 'Dazi,' such as active user numbers, paid conversion rates, and enterprise client renewal rates. Without concrete operational data, the external world cannot accurately gauge the product's potential to transition from a 'free trial' phase to generating actual revenue.

Particularly in the B-end market, renewal rates are a vital metric for assessing product value and user loyalty. The prolonged absence of such data casts doubt on its profitability.

The current competition among AI large models has shifted towards practical applications, with market attention moving from model parameter performance to verifying sustainable business models.

Without reliable financial and operational data as a reference, whether Baidu Dazi can progress beyond the 'showcasing prowess' stage and achieve a sustainable commercial closed loop remains a key point of market observation.

Intensifying Competition: ByteDance and Kuaishou Step Up Their Game, Setting the Stage for Direct Clashes

Following the excitement of WAIC, the AI video generation arena has entered a more intense phase of existing market competition.

Leading companies in the arena have begun deploying strategies along their respective routes, with direct competition between ByteDance and Kuaishou in full swing. In this fiercely competitive market environment, the competitive landscape faced by Baidu Dazi is poised to change.

ByteDance has gained a competitive edge in the short-video intelligent agent field through a dual-wheel architecture of 'model capability + ecological distribution'.

The company has progressively integrated its Seedance model into major ByteDance products like Jimeng, Jianying, Xiaoyunque, and Volcano Engine API, linking the entire process from content production and processing to distribution and monetization. This tightly integrates the underlying model with upper-layer applications, creating extensive scenario coverage and boosting user loyalty in the C-end market.

Based on actual market performance, Seedance accounts for over 80% of daily Token consumption and exhibits strong ecological aggregation effects.

Kuaishou Kling, on the other hand, has established a closed-loop, dual-wheel drive of 'B-end API + P-end subscription', helping Kuaishou secure differentiated advantages in the professional market and B-end commercialization.

On one hand, Kling provides technical support to enterprise clients through API services; on the other, it offers a paid membership system for professional creators (P-end). Currently applied in multiple fields such as advertising, short film production, and game content creation, Kling has demonstrated good user retention on both B-end and P-end.

According to its Q1 2026 financial report, Kling AI generated over 650 million yuan in revenue; by March 2026, its annualized revenue run rate approached 500 million USD.

Baidu Dazi now faces direct competition on two fronts: on the C-end, it must contend with ByteDance's ability to attract users through its ecological closed loop; on the B-end, it faces stiff competition from Kuaishou's commercialization efforts in specialized fields.

Overall, Baidu Dazi's primary challenge lies in the speed of building ecological barriers. If it relies solely on the data and traffic advantages accumulated through traditional search engines without forming unique competitive advantages in application scenarios, developer ecosystems, or business models for video intelligent agents, maintaining its current leading position in the arena will face significant uncertainty.

The Path to Breakthrough: Three Shortcomings That Define Long-Term Competitive Limits

In recent years, Baidu's value proposition has fundamentally shifted from its former 'search advertising' business to a growth model driven by AI technology.

Financial data best illustrates this strategic transformation: in 2025, Baidu's total revenue reached 129.1 billion yuan, with AI business revenue hitting 40 billion yuan. In 2026, this trend accelerated further, with AI business revenue accounting for over 50% of total revenue for the first time, becoming the core engine driving the company's development.

Revenue growth is underpinned by a three-tier revenue structure: the first tier comprises AI cloud computing services, providing a solid computational foundation, with Baidu Intelligent Cloud ranking first in China's AI public cloud market share for six consecutive years; the second tier accelerates AI application innovation, with products like Baidu Wenku and Baidu Netdisk generating stable revenue through high-stickiness subscription models; the third tier encompasses AI-native marketing services, leveraging digital humans to enhance production efficiency and marketing effectiveness.

Above this revenue structure, Baidu has made a significant bet on 'intelligent agents' at the application layer, guiding AI from simple 'content generation' to deep 'execution and collaboration'. As the vanguard product of this strategy, 'Baidu Dazi' represents its intelligent agent tailored for daily productivity scenarios.

In the current market environment, where praise and criticism coexist, 'Baidu Dazi' must overcome several hurdles to achieve long-term development.

The first obstacle is breaking down data silos and establishing a transparent commercial return mechanism. Currently, AI applications widely suffer from 'acclaim without profitability'. In this context, Baidu needs to address external doubts about its commercialization capabilities with real data and quantifiable growth figures, proving that AI applications can not only generate traffic but also deliver tangible economic benefits.

The second challenge is optimizing the technological foundation. The essence of intelligent agents is 'execution', demanding high system stability and response speed. Baidu must tackle technical bottlenecks under high concurrency and ensure users experience smooth, uninterrupted processes when completing complex tasks, preventing them from losing confidence and patience due to technical glitches.

The third hurdle is differentiating itself and avoiding getting trapped in homogeneous competition. Currently, major internet companies are launching AI assistants, making functional convergence an industry norm. Understanding Chinese work habits and cultural backgrounds, transforming cold tools into three-dimensional, warm, and personalized work companions, will be key to standing out in this intelligent agent melee.

On the surface, these three hurdles are product-related, but they fundamentally determine competitive limits. The commercial closed loop determines its development prospects, technological foundation determines its stability, and differentiated positioning determines whether it will be overshadowed. Only by overcoming these three hurdles can 'Baidu Dazi' become a truly indispensable partner for users.