Qwen is one of the strongest AI model families to come out of China. It is also, as of today, the hardest one to use as a normal person in the US. There is no Qwen app in the American App Store. Not a bad one, not a regional one. None.
Search the store for “qwen” and you get DeepSeek, a third-party app called Gwen AI, and ChatGPT. Search “alibaba qwen” and you get Alibaba’s wholesale trade app. The model that wins benchmark tables is not in the store that serves your phone.
I checked the App Store listings and the official pages on September 29, 2026. What follows is where Qwen actually lives for someone who is not a developer, and why the search results are misleading you.
Last updated: September 29, 2026. Model names and hosted versions move fast in this category. Treat anything version-specific on this page as a snapshot.
Qwen is a model, not a product
This is the thing to understand before anything else, because it explains every confusing search result after it.
Most Chinese AI names you have read about are products. DeepSeek is an app you install. Kimi is an app you install. Qwen is a family of models that other people build products from. Alibaba trains it and then publishes it, and the consuming public is not the primary customer.
That shows up directly in what people search for. The autocomplete for “qwen” is a wall of developer intent: api key, coder, gguf, quantised builds, embeddings, fine-tuning templates, hardware requirements, and long strings of model file names. Those are people picking a model to run, not people looking for something to chat with.
Where you can actually use it
Strip out the developer routes and Qwen has a thin consumer surface. It is thin, but it is not empty, and one of the entry points requires no install at all.
| Way in | What it is | Good for |
|---|---|---|
| The web chat | Qwen’s hosted chat interface | The only route with no install. Open it and type |
| The API | Pay-as-you-go access for code | Developers wiring it into something else |
| Open weights | The models themselves, downloadable | Running it on your own hardware. Not a consumer path |
| Third-party hosts | Other companies serving Qwen models | When you want the model without Alibaba’s front end |
| A Qwen phone app | Does not exist in the US store | Nothing. This is the row people are looking for |
That last row is the answer to a search that shows up in volume. If you are on an iPhone or Android in the US and you want Qwen on your home screen, there is no official build to install. What you will find under that name in the store is unaffiliated software riding the term.
What the free tier actually is
“Qwen free tier” and “qwen free api” are both live search terms, which tells you the honest state of things: people are not sure what is free and what is metered, and the answer differs by route.
The web chat is the route where the word free means what you expect. You open it, you type, you are not asked for a card.
The API is the opposite. It is a metered developer product, and the free allowances attached to it are starter credit, not a permanent tier. That distinction gets lost in roundups that list the two side by side as if they were the same offer.
| Route | What free means there |
|---|---|
| Web chat | Free to use. No card |
| API | Metered. Free allowances are starter credit, not a plan |
| Self-hosted | Free to run, and you pay for the machine instead |
| Third-party hosts | Priced by whoever is hosting it |
For a subscription comparison, the useful framing is that Qwen does not sell a consumer subscription in the US the way its peers do. There is no monthly plan to sign up for and no cancellation flow to worry about. That is unusual, and for a lot of people it is the reason to prefer it.
What it is actually good at
Qwen’s consumer-visible strengths are not in chat. They are in the media tools, and they show up in search before they show up in reviews.
| What people search for | What it maps to |
|---|---|
| qwen image | Image generation and, more usefully, image editing |
| qwen video generation | Text-to-video output |
| qwen tts | Text-to-speech, including voice cloning |
| qwen omni | Handling audio and vision in the same conversation |
| qwen vs deepseek | People comparing models before picking a chat front end |
But the comparisons are the interesting column. “qwen vs deepseek”, “qwen vs claude”, “qwen vs kimi” and “qwen vs gemma” all pull traffic. Those searches are not being made by people choosing an API. They are being made by people who want to know which one to open in a browser tab.
And that is the real consumer question. If you are not deploying anything, the model family matters less than the interface you end up typing into.
Where people get stuck
The friction around Qwen is not technical. It is navigational, and it comes from the same root as everything above.
| Symptom | What is going on |
|---|---|
| Searching the app store and finding nothing official | Correct. No consumer app is published there |
| Installing something called Qwen that feels wrong | Probably a third-party app. Check the publisher before you type anything into it |
| Not sure if you can use it without paying | The web chat is the free route. The API is a separate, metered product |
| Confused by model version names | Those names are for people deploying the model, not for users of it |
| Wondering which company is behind it | Alibaba. “qwen which company” and “qwen who owns” are both live searches |
That fourth row is the one I would change if I could. The version numbers are load-bearing for developers and pure noise for everyone else. If you are a normal user, the name of the model you are talking to is not a fact you need.
What I would do
If you want to try Qwen and you are not a developer, the path is short.
| If you | Do this |
|---|---|
| Just want to talk to it | Use the web chat. Skip the store entirely |
| Want it on your phone | Add the web version to your home screen. That is as close as it gets today |
| Want image or video output | Start in the web interface before looking at any paid host |
| Are comparing it to the other Chinese models | Compare interfaces, not benchmarks. That is the decision you are actually making |
This is the reverse of the pattern on most of this site. DeepSeek and CapCut both have real apps and complicated billing. Qwen has neither an app nor a subscription, and the difficulty is entirely in finding that out.
Where it fits next to the rest: the ten-app availability check puts Qwen’s consumer surface next to the ones that do ship apps, the DeepSeek page covers the opposite case where an app exists and the billing is the trap, and the China Tech in America index holds the whole set.
What I would watch is whether Alibaba ships a consumer client in the US. The search demand for one is already there and it currently has nowhere to land. If an official app appears, this page becomes wrong in the best way, and the store listing will be the place to check before trusting anything here.