Wen Xiaoyan

What is Wenxin?

Wenxin is a versatile AI assistant under Baidu; it is also the successor brand to “Wenxin Yiyan” and “Wen Xiaoyan”. The product was launched in 2023 under the name Wenxin Yiyan, was renamed Wen Xiaoyan in 2024, and then renamed Wenxin at the end of 2025.

The user’s existing account and its main functions remain unchanged; therefore, when searching for “Wen Xiaoyan’s official website” or “Wen Xin Yi Yan”, one should be directed to the Wen Xin product, rather than having to look for three separate applications.

Wenxin is based on Chinese-language Q&A, Baidu search, and the Wenxin large model; it now includes features such as task management, multi-modal search, Safe Writing, AI-powered PPT creation, image editing, comic creation, video generation, and video calls in dialects, covering areas such as search, work, learning, and content creation.

Core functions

• Task hosting:The user only needs to describe the goal; the system will automatically plan a multi-step process, search for relevant information, analyze the data, and deliver reports, PPTs, tables, or visual representations.
• All-in-one search:It supports queries via text, voice, images, videos, and documents; it also provides access to map navigation, encyclopedic information, movies and TV shows, tables, mind maps, and health-related data.
• Write with confidence:Generate and refine Chinese content such as reports, speeches, work summaries, lesson plans, novels, etc., and use searches to verify facts and include citations.
• Document Q&A:After uploading a file, it is possible to summarize, extract information, and ask follow-up questions; moreover, files can be synchronized from WeChat chat records, thereby reducing the steps required to find and download them on mobile devices.
• AI PPT：An outline and a complete presentation can be generated from a topic or reference document; it is possible to specify the target audience, length, language (Chinese or English), as well as the template, and then continue to modify the pages before exporting them.
• Magical creation:It supports text-to-image generation, one-click photo editing, styled photos, multi-page comics, bringing static images to life, and creative templates.
• Video generation:Create short videos using a single sentence, one image, multiple images, or the first and last frames; the clarity, duration, and number of videos are determined by the energy consumption.
• Video call:Real-time communication is possible via camera and voice; it can recognize dialects such as Henan dialect, Northeastern Chinese dialect, and Cantonese, and respond in those dialects.
• Multimodal models:The Wenxin 5.0 series can understand text, images, audio, and video, and is used for creative writing, logical reasoning, and cross-modal generation.

Free use and prices for energy points

The Wenxin app can be downloaded and installed for free, and it offers basic Q&A functions, search capabilities, as well as some creative tools. Advanced features for handling images, videos, PPTs, or tasks may require “energy points”.

The current payment options listed on the App Store in China are:

• 1 yuan buys an equivalent small amount of energy.
• 5 yuan, 10 yuan, and 50 yuan tiers
• 100-yuan range
• 500 yuan range

The App Store only shows the amount spent on top-ups, and it does not provide clear information on the specific energy consumption associated with each feature on its public pages. Different models, video lengths, resolutions, the number of PPT slides, and the complexity of tasks can all result in varying levels of energy consumption.

Within the app, the actual deduction will be displayed before execution or on the relevant function page.

Energy top-ups do not constitute a traditional unlimited membership; purchasing 500 yuan does not guarantee unlimited use. The rules regarding gifts offered as part of promotions, expiration dates, refunds, and cross-device balance are subject to the guidelines on the account page.

Baidu Qianfan API and costs

The Wenxin Model API, designed for developers and enterprises, is made available through Baidu Qianfan. Qianfan currently builds a platform based on model services, Agent engines, tools and MCP, knowledge bases, fine-tuning, evaluation, agent sandboxes, and enterprise governance; its interfaces are compatible with the common Chat Completions format.

The models include the ERNIE 5.0 and ERNIE 4.5 series, as well as various third-party text, multimodal, image, and video models.

Language models are charged based on the input, output, and cached tokens, while video generation is billed according to a point system; online searches incur search requests as well as additional input tokens. According to the information in this document, an online search costs around 0.004 yuan per time, and video generation is priced at 1 yuan per point, with different specifications requiring different numbers of points.

The model rates and time-limited discounts change frequently; the final amounts are determined according to the Qianfan billing page and the order details. Newly registered users who have completed identity verification can receive a voucher worth 20 yuan, valid for one month.

The energy level in the Wenxin APP is not interconnected with Qianfan vouchers or the API balance.

Open-source status

In June 2025, Baidu made some of the weights related to the ERNIE 4.5 series available, providing versions in Pytorch and Paddle formats; the specific dimensions, models, and licenses need to be checked individually.

Baidu also maintains projects related to PaddlePaddle and ERNIE, as well as the Qianfan SDK on GitHub; some of the code in the Qianfan SDK is licensed under Apache 2.0.

This does not mean that all models of Wenxin 5.0, the Wenxin app, Baidu Search, task hosting services, and the Qianfan Cloud platform are fully open-source. The consumer-facing products should be labeled as proprietary services in the catalog, with a note stating that “some weights and SDKs related to ERNIE 4.5 are open-source”; it cannot simply be stated that “Wenxin is open-source”.

Platforms and privacy

Wenxin is available for Web, Windows, macOS, Android, iPhone, and iPad. For the Apple version in China, iOS or iPadOS 13.0 or later is required; Macs with Apple chips can run the corresponding version as well.

Task hosting and new models may be made available in phases, and the features of different platforms are not always synchronized.

Voice, video, location data, images, documents, and files synchronized via WeChat are processed for upload based on their specific functions. Medical reports, identity information, contracts, and company-related documents should be used only after obtaining authorization, and their sensitive data must be removed first.

Video calls and responses in dialects enhance the convenience of interaction, which means that camera, microphone, and location permissions should be enabled as needed.

Wenxin (original title: Xiaoyan) Usage Guide

Complete a basic task.

1. Clarify the issue, time frame, location, source priority, and output format;
2. Upload materials for which usage rights exist, or enter search queries in Wenxin (the original text).
3. First, establish a framework using task hosting, and then use universal search to gather additional evidence.
4. It is necessary to distinguish between factual information from the source, the author’s opinions, and AI-generated conclusions.
5. Check each item for dates, numbers, the original location, and any conflicting evidence;
6. The conclusions are manually revised, the verification time is recorded, and then they are published;

Create reusable professional workflows

1. Break down complex topics into four categories of questions: background, data, comparison, and conclusions;
2. Task orchestration, comprehensive search, and secure writing constitute the fixed research steps;
3. Give priority to using the official website, research papers, regulatory documents, and raw data;
4. A second person is assigned to review conclusions that are considered high-risk;
5. Save queries, evidence, versions, and unresolved issues;
6. Re-run after the data changes and update the conclusions;

Which users are it suitable for

• Ordinary users and those in the workplace who need to use Baidu for searches, write in Chinese, and verify information.
• Office staff who wish to handle tasks such as report preparation, PPT creation, and data organization
• Creators who produce images, comics, edit photos, and make short videos
• Enterprise developers who access the ERNIE model and Agent capabilities through Baidu Qianfan

Reliability and limitations

• “Write with confidence” and search for references do not guarantee that the content is absolutely accurate;
• AI may misinterpret web pages, invent numbers, or generate fake references; medical, legal, financial, and policy-related information should be verified against authoritative sources.
• Creating images and videos also involves considerations regarding the rights to upload materials, portraits of people, trademarks, music, and training content; manual review is required before such content can be used for commercial purposes.

Frequently Asked Questions

Why can’t Wen Xiaoyan be found?

Wen Xiaoyan has been renamed Wen Xin once again; Wen Xin Yiyan, Wen Xiaoyan, and the current Wen Xin all represent different stages in the evolution of the same product.

Is Wenxin free?

Basic Q&A and some features are available for free; advanced generation capabilities may require energy points, with top-up options ranging from 1 to 500 yuan.

Does purchasing energy points include an API?

It is not included. The energy value in the Wenxin APP is used for functions on the consumer side, while the Qianfan API requires separate activation and is paid via tokens, points, or resource packages.