Qoder and Qwen Code foundations merged into one
China's Alibaba on Nov. 11 released "Qwen Code 1.0 Preview," the next version of its AI programming agent "Qwen Code," as a test release on GitHub. It merged the agent foundations that had been developed separately for its commercial AI coding service "Qoder" and for Qwen Code into one. The merged foundation was released as open source, available for anyone to use.
An agent foundation, also called a "harness," is the mechanism that has an AI model read and write files and run commands to move work forward. Qwen Code is an AI coding tool that runs in the terminal (the command-line screen), and Alibaba released it in July 2025 together with "Qwen3-Coder," a model for programming. Qoder is a service with paid plans that the company launched in August 2025, offering a dedicated integrated development environment (IDE) and a CLI.
According to GitHub records, version "v1.0.0-preview" was published at 12:55 p.m. Nov. 11 Japan time. It is labeled a "Pre-release," ahead of the official version, while "v0.27.2" in the existing line was released in the early hours of the same day, Nov. 11 (Japan time). 1.0 Preview is distributed separately from the existing versions.
Connects to other companies' models, installs without Node.js
According to the release notes, 1.0 Preview takes instructions in natural language to understand code, implement features, investigate bugs and run tests. It can be used with Qwen and Qoder accounts and can also connect to multiple model providers and to users' own endpoints with compatible formats. It supports resuming conversations closed midway through work and running without human operation, so it can be built into scripts and automation.
The official documentation lists DeepSeek, Grok, MiniMax, Z.AI, Kimi and OpenRouter, among others, as other providers. Supported connection formats are OpenAI-compatible, Anthropic, Google's Gemini and Vertex AI.
Installation has also been simplified. Packages for macOS, Linux and Windows run without separately installing Node.js, the JavaScript runtime. Installing through npm, a Node.js tool, requires Node.js 22 or later.
The license is Apache 2.0, which permits commercial use and modification.
Test release, with no pricing or default model given
Alibaba said the interface, settings and behavior may change during the preview period. The README describes 1.0 as being in the preview stage and changing rapidly.
On the day of release, a 1.0.0-preview user reported on GitHub that the connection to the model repeatedly dropped during runs. The cause is unknown, and the developers' automated triage has asked for additional information such as logs (as of Nov. 11).
Eligibility depends on what users connect to. The documentation says using a Qwen account requires a Plus or Pro subscription to chat.qwen.ai. Pricing and usage limits are not stated in the documentation. The README says available models and pricing depend on the provider and account. Pricing and limits when using a Qoder account, and the model used by default, are not given in official materials.
No performance comparison yet
Terminal-based coding agents are a field where Anthropic's Claude Code, OpenAI's Codex and Google's Gemini CLI, among others, compete. Alibaba has moved the foundation it had used in its paid service to the open-source side and made it able to connect to other companies' models. Qwen Code's GitHub repository has about 29,400 stars (registrations showing interest) (as of Nov. 11).
Alibaba did not give performance figures in the release. No third-party evaluation comparing 1.0 Preview with other companies' tools under the same conditions could be found as of Nov. 11.
