Qwen Code stopped working: why and what to do in 2026

The query “Qwen stopped working” can mean several different failures, each with a different solution. First, you need to identify exactly what is failing.

Below are the failures in order: what happened to free sign-in, how browser chat differs from Qwen Code in the terminal, why the agent launches but returns an error, and how to get everything working again without a card. Everything was checked against Qwen's official documentation as of September 2, 2026.

Information about working with Ailoot is available in the documentation.

What does “Qwen stopped working” mean?#

First, it’s worth understanding exactly what broke, because people use “doesn’t work” to describe four different things. One person can’t sign in, another’s agent launches but returns an error, a third can no longer open browser chat, and a fourth has run out of quota. Each has its own fix, and if you start in the wrong place, you’ll lose your evening.

It’s important here to separate two questions that get conflated in complaints. First: does the request reach the server? Second: does the server let you in without the right access?

The simple takeaway to start with: don’t guess about VPNs and blocks until you’ve checked sign-in and the model. In most cases, Qwen “stopped working” precisely because Qwen’s developers changed the access rules for everyone at once, while the old settings on your computer stayed the same.

Where did free sign-in to Qwen go?#

The most common reason something “stopped working” is that free sign-in was shut down. Before, when you launched Qwen Code for the first time, it opened sign-in through a browser; you signed in to your qwen.ai account, and the agent got a free quota of around two thousand requests per day. Many guides in search results still describe exactly this setup.

That setup is gone. The official sign-in documentation says, verbatim: “its free tier was discontinued on 2026-04-15,” meaning free access ended on April 15, 2026. It also says that Qwen OAuth sign-in was removed from the sign-in method selection dialog and remains in the documentation only as a disabled option. This was Qwen developers’ decision for everyone at once, and it doesn’t depend on the client’s previous local settings.

So if it used to “just work without a card” and then suddenly stopped, look for the cause in the shutdown, not in a block: the old free sign-in is gone, and you need to replace it with another access method. Which one? More on that below, including a free option.

Qwen in the browser and Qwen Code in the terminal: which one stopped working?#

People use the word “Qwen” to mean two completely different things, and that confusion is another source of “stopped working” complaints.

The first is browser chat on chat.qwen.ai. It’s a regular chat: you visit the page, type a question, and get a text response. You copy code from it by hand. If this is what “stopped working,” the issue is almost always signing in to the site itself or regional restrictions on the page, and the fix is on the browser side.

The second is Qwen Code, a separate tool installed in the terminal that works as an AI agent. According to its official description, it’s an open coding agent that reads the whole project, opens files on its own, edits them, and runs commands. It doesn’t respond with text in a window; it works inside the folder containing the code. And when it “stops working,” it means something else: most often, the connection to the model has failed.

First, figure out which one you were using. The rest of the article is mainly about Qwen Code, because “it wrote code by itself and stopped” is about the agent in the terminal. Browser chat doesn’t work that way at all.

Why does Qwen Code launch but return an error?#

A common scenario: Qwen Code installs, the qwen command runs, the window opens, but as soon as you give it a task, you get an access error, usually code 401, which in server language means “access not confirmed.” The agent itself is fine; the problem is which model it’s connecting to.

The cause is again the shutdown of free sign-in. Before, the agent connected to the model by default through free qwen.ai sign-in. Now that sign-in has been removed, the setting on your computer still points the old way—and that’s why you get an error. Until you specify a new access method, the agent has nowhere to go for an answer.

According to the official documentation, there are now three methods. The first is an Alibaba Cloud plan: Coding Plan, with a weekly quota for one person; Token Plan, with pay-as-you-go pricing for teams; or a regular Model Studio key. The second is a third-party provider’s key: Qwen Code can work with DeepSeek, Kimi, MiniMax, ModelScope, and others; the key is entered in the settings file. The third is your own model on your own computer, without any key. The last option is free, and has its own section below.

How can you get Qwen working again for free and without a card?#

If your goal is to get Qwen working again without spending money or using a card, there’s one working way: run the model on your own computer. Qwen Code is designed to be flexible for this. Its description explicitly says it supports any third-party provider or local model through Ollama and vLLM, and that you can switch between them on the fly.

The idea is simple. Install a program on your computer that can run models locally (Ollama is the easiest option), download a Qwen model through it, and it will start a local server on your machine. Then, in Qwen Code’s settings, choose “custom provider” and enter the address of that local server. Now the agent sends requests directly to the model on your computer, without going through the cloud.

Steps:

1. Install Ollama from the official website, ollama.com—it’s available for Windows, macOS, and Linux.
2. Download a Qwen model of the desired size through Ollama (the larger the model, the more powerful your computer needs to be).
3. Start Ollama—it will start a local server on your computer.
4. In Qwen Code, open the access settings with the /auth command and choose “Custom Provider.”
5. Enter the address of the local Ollama server as a compatible endpoint, and the agent will start working through it.

There’s one real cost to this option: hardware. The model runs on your computer, so a weak laptop won’t handle the heavier versions; you’ll have to choose a smaller model. But there’s no sign-in, key, or card: the computations run locally, without contacting a cloud model.

How can you connect Qwen through a third-party provider if your computer can’t handle a local model?#

When your own computer can’t handle it, the cloud is still an option, and it doesn’t have to be through Qwen itself. The agent can connect to other providers’ models using a key; this is a standard feature, with no tricks involved.

According to the documentation, Qwen Code works with keys from different providers: DeepSeek, Kimi, MiniMax, Z.AI, ModelScope, and others. The process is the same for all of them: set up access with the provider you choose, get a key, enter it in the ~/.qwen/settings.json settings file, and specify the model name and service address there as well. After that, the agent gets its responses from that provider. It no longer connects through Qwen’s disabled free sign-in. For information on issuing a key and what it costs to use one, see API key for Qwen Code.

A note about Qwen itself: Alibaba Cloud has a plan for developers with a weekly quota, as well as a regular Model Studio key. Network access to the server and payment for access are separate questions; don’t conflate them when diagnosing the problem.

What should you do step by step if Qwen stopped responding?#

Here it is all in one sequence so you don’t have to poke around at random. Go from top to bottom, stopping as soon as you find the answer.

1. Identify the product. Browser chat on the website and Qwen Code in the terminal are different things with different failure modes.
2. Remember how you signed in. If free sign-in without a card used to work, it was shut down on April 15, 2026; that’s the cause, and you’ll need a new access method.
3. Read the error message. Code 401 and wording about access mean the model isn’t connected. It says nothing about a block.
4. Choose based on your hardware. If your computer is powerful, run a local model through Ollama—it’s free and doesn’t require a card. If it’s weak, connect to a model using a third-party provider’s key.
5. Check network access separately, after checking the product, sign-in method, and model connection.

That’s how “Qwen stopped working” breaks down into clear steps, and in nearly every case the cause turns out to be a change in access rules. The country only comes into it last.

Get to know Ailoot.