Qwen logo, a blue-violet hexagonal geometric emblem, for a guide to Qwen Code 1.0 Preview

Qwen Code 1.0 Preview: Power-User Guide to Install, Auth and Qoder Features

by

Last updated: November 12, 2026

Qwen Code 1.0 Preview, released on November 11, 2026, merges the Qoder agent harness into Qwen’s open-source terminal coding agent. It installs as a native package on macOS, Linux and Windows without Node.js, signs in with Qwen or Qoder accounts or any compatible endpoint, and supports session resumption and non-interactive runs.

Contents

This guide uses the official v1.0.0-preview release notes, the v1.0.0-preview.1 notes, the 1.0 Preview README and the Qwen Code authentication docs. If you used the earlier CLI, our Qwen Code CLI 0.26 guide covers the October releases.

What is new in Qwen Code 1.0 Preview

Qwen Code is an open-source coding agent that runs in your terminal. You describe a task in plain language, and it reads code, makes changes, investigates problems and runs tests.

The v1.0.0-preview release lists four headline changes:

  • One agent, two harnesses. It brings together the agent capabilities of Qoder and Qwen Code for code understanding, implementation, debugging and testing.
  • More ways to connect a model. It supports Qwen and Qoder accounts, multiple model providers and custom compatible endpoints.
  • Native packages. It ships for macOS, Linux and Windows without a separate Node.js installation.
  • Long-running and scripted work. It supports session resumption and non-interactive execution for ongoing development, scripting and automation.

A day later, v1.0.0-preview.1 fixed Qwen subscription accounts that could not use subscription models after sign-in. The model catalog now follows your sign-in state automatically. It also synced the subscription model catalog and thinking-budget mapping with Qwen Code 0.27.x, and Qoder models now support parallel tool calls, so multi-tool tasks no longer run strictly one after another.

How to install and start Qwen Code 1.0

The README gives a one-line installer for each platform. You can also download a package for your operating system and architecture from the release assets.

  1. On Linux or macOS, run:
    curl -fsSL 'https://qwen-code.oss-accelerate.aliyuncs.com/install.sh' | bash
  2. On Windows, run this in PowerShell:
    irm https://qwen-code.oss-accelerate.aliyuncs.com/install.ps1 | iex
  3. Open a new terminal, go to your project folder with cd my-project, and run qwen.
  4. Follow the connection dialog at startup, or type /auth during a session to connect a model.
  5. Type /model to choose from the models your account or key can use.
  6. Start with a read-only request, such as asking it to explain how the project starts and where tests live, before asking for changes.
  7. Review each permission prompt, then use /diff to inspect changes and check the reported test results before you accept them.

Sign-in and model options

The README groups connections into three types, and the authentication docs (last updated November 9, 2026) give the details.

Connection What you need Notes from the docs
Qwen account Plus or Pro subscription from chat.qwen.ai Browser sign-in through Qwen OAuth; not usable in CI or unattended containers
Qoder account A Qoder account Listed as a sign-in entry in 1.0 Preview; preview.1 added parallel tool calls for Qoder models
QwenCloud Token Plan key or pay-as-you-go API key China and International endpoints; enables the built-in web_search tool
Alibaba Cloud Model Studio Coding Plan, Token Plan or API key Coding Plan includes a weekly quota and uses a dedicated endpoint
Third-party provider Provider API key Built-in options include DeepSeek, Grok, MiniMax, Z.AI, Kimi, ModelScope, OpenRouter and Requesty
Custom provider Endpoint, credentials and model IDs OpenAI-compatible, Anthropic, Gemini and Vertex AI protocols

The README states that model access and service charges depend on the provider and account you choose. The tool itself is free under the Apache License 2.0. For plan background, see Is Qwen free? and our Qwen4-Max-Preview guide.

The standalone qwen auth command has been removed. The docs say to run /doctor inside Qwen Code to check your current authentication.

Power-user workflows

Resume work across days. Run qwen --continue to reopen the most recent saved session for the same project, or qwen --resume to pick an earlier one.

Script it. For a single request without the interactive interface, pass a prompt with -p. Because the Qwen account sign-in needs a browser, the docs recommend an API key for CI and containers. A headless setup based on the documented variables looks like this:

# Pay-as-you-go QwenCloud key (International)
export QWENCLOUD_API_KEY="sk-your-api-key"
export OPENAI_BASE_URL="https://maas.qwencloudapi.com/compatible-mode/v1"
export OPENAI_MODEL="qwen3.8-max"
qwen -p "Explain the purpose of this repository. Do not modify files."

The README advises configuring authentication and permissions before any unattended run.

Keep prompts specific. The README’s sample prompts follow a useful pattern: ask the agent to read the relevant files and cite their paths, then request a focused change that follows existing patterns and runs the tests that cover it.

Switch models per terminal. The docs show qwen --model "qwen3-coder-plus" in one terminal and a different model in another, which helps when you compare a fast model with a stronger one on the same repository.

Qwen Code 0.27 vs. 1.0 Preview

The 0.x line continues in parallel. Version 0.27.2 shipped on November 10, and the November 5 weekly update still tells users to upgrade with npm i @qwen-code/qwen-code@latest -g.

Area Qwen Code 0.x (0.26–0.27) Qwen Code 1.0 Preview
Install npm package Native packages and install scripts; no separate Node.js
Agent harness Qwen Code Qwen Code plus Qoder agent capabilities
Accounts Qwen account, QwenCloud, Model Studio, providers Adds Qoder accounts
Status Stable releases Preview; behavior may change

Compared with other open-source terminal agents, such as OpenCode 2.2 and the Pi coding agent 1.5, Qwen Code’s main draw is first-party access to Qwen and Alibaba Cloud plans alongside third-party and custom endpoints.

Limitations and preview caveats

  • The Qwen team says some interfaces, configuration options and behaviors may change before 1.0 stabilizes, so read the release notes before each upgrade.
  • Community pull requests are paused; the team expects to reopen contributions after December 2026. Issues and suggestions are still accepted.
  • Qwen account sign-in requires a paid Plus or Pro subscription and an interactive browser flow.
  • The release notes do not publish a context window or usage limits for the agent itself; those depend on the model and plan you connect.

Qwen Code was originally based on Google Gemini CLI v0.8.2 and has been developed independently since v0.1, according to the README.

Frequently asked questions

What is Qwen Code 1.0 Preview?

Qwen Code 1.0 Preview is the first 1.0 release of Alibaba’s open-source terminal coding agent. Released on GitHub on November 11, 2026, it combines the agent capabilities of Qoder and Qwen Code for understanding, implementing, debugging and testing code, and ships as native packages for macOS, Linux and Windows.

Do I need Node.js to install Qwen Code 1.0?

No. The 1.0 Preview release notes say it provides native packages for macOS, Linux and Windows without a separate Node.js installation. You can install it with the official shell script on Linux and macOS or the PowerShell script on Windows, or download a package from the GitHub release assets.

Is Qwen Code 1.0 free?

The code is free and open source under the Apache License 2.0, but model use is not automatically free. The README says model access and charges depend on the provider and account you connect. Signing in with a Qwen account requires a Plus or Pro subscription from chat.qwen.ai, according to the Qwen Code authentication docs.

Which models can Qwen Code 1.0 use?

You can connect a Qwen or Qoder account, QwenCloud or Alibaba Cloud Model Studio plans and API keys, built-in third-party providers, or a custom compatible endpoint. The docs list OpenAI-compatible, Anthropic, Google Gemini and Vertex AI protocols. Use /model inside a session to pick a model.

Is Qwen Code 1.0 stable enough for daily work?

It is labeled Preview. The Qwen team says interfaces, configuration options and behaviors may change before the release stabilizes, and asks users to read the release notes before upgrading. The team is also not accepting community pull requests until after December 2026.

Sources

Related answers

Leave a Comment