Ant Group’s LingGuang surges past one million downloads in four days

A viral AI that builds apps from a sentence

Ant Group’s new app LingGuang moved from debut to mass adoption in days. The artificial intelligence assistant crossed one million downloads within four days of launch in China, according to the company. Traffic spiked so sharply that its core make an app feature was briefly suspended on Thursday evening while engineers added capacity. The service returned soon after and the surge continued.

The speed stands out in a crowded field of AI tools. Ant Group says LingGuang reached its first million faster than ChatGPT and Sora did after their initial releases. By the sixth day, downloads had climbed past two million. On Apple’s mainland China App Store the app reached the top spot among free utilities by Saturday and ranked seventh across all free apps, then rose to sixth early the next week. The product is available on the Apple App Store, major Android stores, and the web for users in and outside China.

What is vibe coding and why this approach matters

Vibe coding is a new way to build software. Instead of writing code, a user describes the goal in plain language and the AI does the heavy lifting. Many tools today suggest or write code snippets that a developer still needs to review and integrate. LingGuang goes a step further. It generates a working mini app that runs immediately, then stays in the chat to revise the app based on feedback.

The appeal is clear. People who have ideas for a small tool, a calculator, a planner, or a game can test those ideas without hiring a developer or learning a programming language. The broader market for this kind of tool is growing fast. US based Replit, a leader in vibe coding, reported annual recurring revenue of about 100 million dollars in mid 2025, up from around 10 million at the end of 2024. Ant Group says LingGuang users typically go through about six rounds of edits per session as they refine their first draft into something they like.

Inside the app features

LingGuang is framed as a multimodal assistant. It understands and produces text, images, audio, and data. The app can respond with 3D models, interactive charts, animations, maps, and small interactive programs. Ant Group says it uses a system of collaborating agents so that different skills can work in parallel and the user sees a single, coherent result.

Flash App, build in 30 seconds

The headline feature is Flash App. A user types a sentence such as create a weekly fitness planner or I want a recipe generator for low sodium meals. In about 30 seconds the assistant returns a functioning mini app with buttons, fields, and logic. That app can be customized through conversation, shared with friends, and connected to backend services where suitable, such as spreadsheets or simple databases.

Ant Group has highlighted examples from early users, like a kid activity generator, a car cost savings calculator, and training tools for new employees. The company says it had to run multiple rounds of emergency capacity expansion to keep up with demand as people tried to create and share their first apps.

He Zhengyu, Ant Group’s chief technology officer who oversees the project, framed the goal in plain terms.

“LingGuang is bringing every user their own personal AI developer, someone who can code, create visuals, build programs, and turn complex ideas into simple solutions, right in your pocket.”

Multimodal outputs that explain ideas visually

Beyond code, the assistant can explain complex topics with pictures and motion. In a study session it can extract key points, arrange them in a logical outline, and pair the text with 3D illustrations or short animations. In a data task it can read a table and propose an interactive chart that updates as the inputs change.

AGI camera and real time scene understanding

A feature called AGI camera, AGI stands for artificial general intelligence, lets LingGuang analyze scenes through the phone camera and provide context or edits on the fly. A user can ask what is the fastest way to measure the space in this living room or remove the background on this clip and brighten the subject. The same tool can generate images or short videos from text or from an existing photo. Features that process live footage raise privacy questions, so clear controls and on device safeguards will be key.

Adoption metrics and infrastructure strain

Ant Group says LingGuang crossed one million downloads on day four. By day six it exceeded two million. On launch week the app reached first place among free utility apps on Apple’s mainland China App Store and moved from seventh to sixth in the overall free charts over the weekend. The rapid uptake mirrored the strain behind the scenes.

Three days after launch the company temporarily suspended the Flash App function during peak hours while it scaled servers. The team restored the feature the same evening and expanded capacity in several waves after that. Ant executives described a pattern in early behavior. Many people generate a first version, then iterate quickly. Some early users edited their apps more than 100 times in a short window. Average sessions show about six rounds of back and forth before a user is satisfied.

Who built it and what runs under the hood

LingGuang was built by a team inside Ant Group and was supervised by CTO He Zhengyu, a PhD graduate of the Georgia Institute of Technology. The project has been in active development since March 2025. Ant Group has also set up an AGI Lab to push its own large AI models and has launched other AI products this year, including a healthcare assistant and a service robot.

The app can work with several Chinese language models, including Qwen from Alibaba Cloud and selected open source models. Ant executives describe the relationship between LingGuang and Alibaba’s Qwen app as complementary. One focuses on a general chatbot experience at massive scale, the other focuses on building apps and visual outputs on mobile devices.

Cai Wei, LingGuang’s technical lead and a former Google engineer, said the product marks a first in mobile app creation for everyday users.

“It is the industry’s first mobile app that allows users to easily generate and customize apps.”

China’s AI app race and global context

Consumer AI apps in China are moving up store charts at a pace that is rare. Alibaba’s Qwen chatbot crossed about ten million downloads within a week of public testing. ByteDance has pushed its own assistants and media tools into top rankings. During the same period LingGuang climbed into the top tier of the Apple charts. Strong marketing helps, yet the stickiness will depend on how often people return to build or use small tools that solve daily problems.

The model of building from natural language is gaining global traction. Developers and hobbyists use tools like Replit or Copilot to generate code that then gets compiled, tested, and deployed. LingGuang skips straight to a working mini app that lives inside the chat session and on the phone. That removes several steps, lowers friction, and expands the audience to people who have never touched a code editor.

Risks, quality, and safety

Turning a sentence into an app is exciting, yet it brings real risks. Automatically generated code can have bugs or create security gaps if it handles sensitive data. People may publish or share an app that looks polished but fails in edge cases. Clear warnings, permissions, and upgrade paths will matter.

AI output can also be mistaken or biased. If a calculator uses incorrect formulas or a planner gives poor advice, users can make bad decisions. The team will need strong evaluation for tasks that involve health, finance, or education. That could include vetted templates, human review for public galleries, and opt in connections to external data.

Data privacy is another pressure point. Real time camera features process a stream of images and sometimes audio. Users will expect transparency on what is processed on device and what is sent to cloud servers, with options to restrict storage and sharing. China also has strict content and data rules for AI systems. Services that scale fast must prove they can meet those rules while still shipping new features at a steady clip.

Early use cases and how people are testing it

The first wave of LingGuang apps look practical and playful. Parents are building week planners with activity suggestions for kids. Drivers are assembling calculators that compare fuel and maintenance costs for different models. Office teams are spinning up quick training checklists for new hires. Students are experimenting with study guides that mix short text, diagrams, and small quizzes.

In each case the cycle is the same. Describe what you want, check the result, then ask the AI to adjust the layout, add a rule, or connect a data source. The assistant keeps track of the latest version and lets users share a link or QR code. If the editor feels natural, many people who would never open a code repository can still ship something useful to their family, class, or team.

Key Points

• LingGuang surpassed one million downloads in four days and passed two million in six days.
• The Flash App feature briefly went offline during a traffic spike and was restored the same evening.
• It ranked first in free utilities on Apple’s mainland China App Store and moved from seventh to sixth overall over the weekend.
• The assistant generates complete mini apps from plain language prompts, then refines them through chat.
• Features include multimodal outputs like 3D models, charts, animations, and an AGI camera for real time scene understanding.
• Ant Group says average sessions include around six rounds of edits, and some users made over 100 changes to a single app.
• The system uses multiple models, including Alibaba Cloud’s Qwen and open source options, and was built under CTO He Zhengyu.
• LingGuang is available internationally on iOS, Android, and the web.