Baidu said its ERNIE Bot chatbot passed 100 million users on December 28, 2023, about four months after it became publicly available in China. That was a notable early-scale milestone, not a current user count or proof that Baidu had won the AI race. Baidu later announced larger figures, but its public disclosures shifted toward model releases and API activity rather than a comparable, current chatbot-user total.
What ERNIE Bot is—and what the 100 million figure meant
ERNIE is Baidu’s family of foundation models. ERNIE Bot, also known by the Chinese names Wenxin Yiyan and Wenxiaoyan, is the consumer-facing conversational product built around those models. Baidu’s Qianfan platform is a separate part of the business: it gives developers and companies access to models and AI services through Baidu AI Cloud.
Depending on the model and interface, ERNIE products have supported tasks such as text generation, document questions, image understanding, image generation, webpage reading, search and coding. Features can vary by model, account and region; a product name alone does not establish which capabilities a user can access.
Baidu CTO Wang Haifeng announced on December 28, 2023 that ERNIE Bot had surpassed 100 million users. The announcement was reported by Reuters. Baidu did not provide an independently audited methodology in that announcement specifying whether “users” meant cumulative registrations, unique people, active users or another measure. It is therefore best read as a company-reported scale milestone, not a like-for-like comparison with rivals’ monthly active users, downloads or paying customers.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match#1 Best Overall
How ERNIE Bot reached the milestone
Baidu introduced ERNIE Bot in March 2023 through a limited rollout, then made it publicly available in China in August. ERNIE 4.0 followed in October; Baidu said it was comparable to GPT-4, and the model became available to users in November. Those capability comparisons were company claims, not a universal independent ranking. The 100-million announcement came at the end of that first public year. The launch and release chronology was covered by the Associated Press and in its report on public availability.
The milestone mattered because Baidu was among the early Chinese companies seeking to turn the global ChatGPT moment into a domestic consumer and business offering. Its search business, apps, cloud infrastructure and existing AI work gave it potential routes to distribute models beyond a standalone chatbot. But a large early audience cannot, by itself, establish how often people returned, whether they paid, or whether businesses adopted the underlying models.
What the later numbers say—and what they do not
| Date | Reported figure | How to interpret it |
|---|---|---|
| December 28, 2023 | More than 100 million users | Baidu CTO announcement; the definition of “users” was not detailed in the announcement. Reuters report. |
| April 2024 | More than 200 million users | A later Baidu-reported milestone, covered by ThePrint; the report does not make it a verified active-user measure. |
| November 2024 | 340 million users | A figure attributed to Baidu and reported by the Associated Press; the methodology is not established in that report. |
| November 2024 | 1.5 billion daily API calls | Baidu’s Q3 disclosure described API activity, not people or chatbot accounts. Baidu filing. |
| December 2024 | 1.65 billion daily API calls | Baidu’s Q4 results reported calls to its AI models; an API call is not a user. Baidu results. |
The later user figures suggest that Baidu continued reporting growth after the 2023 headline, but they should not be treated as a clean time series of active engagement: the public descriptions do not supply a consistent, audited definition. API calls measure a different kind of activity. They can come from software and business integrations as well as consumer-facing use, and repeated calls do not equal distinct people.
Baidu’s 2026 annual filing reports 679 million monthly active users for the Baidu App in December 2025. That is the broader app, not ERNIE Bot, so it is evidence of potential distribution reach rather than a chatbot audience count. Baidu’s filing does not make the two metrics interchangeable.
From ERNIE 4.0 to ERNIE 5.1
The product did not freeze at the version associated with its first surge in users. In March 2025 Baidu introduced ERNIE 4.5, a multimodal model, and ERNIE X1, a reasoning model. It also said X1 delivered performance comparable to DeepSeek R1 at half the price. That is Baidu’s comparison; it should not be generalized across all workloads without benchmark and pricing conditions. The launch announcement describes the company’s claims and product changes: Baidu’s announcement distributed by PR Newswire.
Baidu said ERNIE Bot became free for PC and mobile users from April 1, 2025. “Free” describes consumer access as announced by Baidu; it does not mean every interface, feature, region or usage level is unrestricted. In June 2025, Baidu released the ERNIE 4.5 series as open-source models. That release concerns the model series; it should not be taken to mean that all training data, training code and processes were made open. Baidu’s ERNIE announcement describes the release.
Rank #3
Baidu’s annual filing says it released an updated ERNIE 5.0 in January 2026, and its Q1 2026 results say ERNIE 5.1 launched in May 2026. These dates show continued product development through the latest disclosures in this article, not a public ranking proving that one model leads every task. The annual filing and Q1 results document the releases.
ERNIE’s competition is not one simple leaderboard
Chinese AI products compete across different audiences and business models. DeepSeek raised pressure around reasoning capability and cost. ByteDance’s Doubao benefits from its own distribution and competes as a consumer assistant. Alibaba’s Qwen combines a model family with developer and open-model activity. Moonshot AI’s Kimi is associated with long-context and research-oriented use cases, while Tencent Yuanbao can draw on Tencent’s ecosystem. Zhipu, MiniMax and other model companies also compete, particularly in enterprise and model services. The AP’s overview of Chinese AI models places ERNIE alongside competitors including DeepSeek and Doubao.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →For users outside mainland China, ChatGPT, Claude, Gemini, Grok and Perplexity are also relevant alternatives. Availability, language performance, data policies and legal or regulatory conditions vary by country. A model’s reputation in one market does not establish that it is accessible or suitable everywhere.
These products should not be ranked as if they competed on a single stable measure. A consumer may care about Chinese-language responses and app convenience; a developer may prioritize API cost, context length, reliability or deployment control. A benchmark result is meaningful only when the task, model version, test method and date are clear.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Where Baidu may have an advantage—and where the evidence is thinner
Distribution and an integrated stack
Baidu can connect models to search, its mobile app, cloud services and productivity products, rather than relying only on a standalone chatbot. For businesses, Qianfan offers a route into Baidu’s cloud and developer platform. Baidu’s consumer product and enterprise platform serve different buyers and have different economics: consumer reach does not automatically translate into API revenue or profitable deployments.
Pressure on engagement and monetization
The available user milestones do not establish retention, frequency of use, willingness to pay or customer satisfaction. Making a consumer chatbot free can lower the barrier to trying it, while leaving monetization dependent on cloud, enterprise, advertising or embedded-product value. The reported API-call volume indicates substantial model activity, but by itself says neither how much revenue it generated nor whether demand is profitable.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Fast-changing capability and market conditions
Model advantages can erode quickly as competitors release new systems, lower prices or open model weights. The 2023 user total cannot answer whether ERNIE 5.1 is preferable for a particular task in 2026. Mainland Chinese services also operate within local regulatory requirements, which can shape product behavior and available answers; international access and usability may differ. Those constraints matter when comparing services across borders.
How to evaluate ERNIE for personal or business use
If you are choosing a chatbot
- Check whether the service is available where you live and whether you can access the specific model you want; the interface may expose different versions or features.
- Test the language and tasks that matter to you. Chinese-language usefulness may not be reflected by English benchmark claims.
- Review the current privacy and data-handling terms before entering confidential or sensitive material.
- Verify important answers independently, especially facts, citations, financial information and code.
- Consider regional behavior and tool access; search, document, image and webpage functions can vary by account, model and location.
If you are evaluating an API or deployment
- Confirm availability in your target region, the endpoint and model version, rate limits, service commitments, context length and supported modalities.
- Compare current input and output pricing for your expected workload rather than relying on a headline price or an old benchmark.
- Assess data retention and training-use terms, Chinese-language quality on your own domain tasks, and migration effort from existing providers.
- Compare hosted inference with self-hosting an open-weight model. Open weights can offer deployment control, but hardware, engineering, evaluation and operations still cost money.
Baidu’s Qianfan platform is the relevant entry point for its enterprise and developer services; its international documentation and pricing page lists model-specific terms that can change. Confirm the current endpoint, region, currency and billing units before committing to a workload.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
