Alibaba's Qwen Code v0.6.0 adds experimental Skills and multi-provider support

If you rely on AI coding assistants inside your terminal or editor, Alibaba's tooling just got a lot more flexible.

The Alibaba Qwen team rolled out Qwen Code v0.6.0, now installable directly via npm.

Multi-provider support and pragmatic polish

The biggest shift in this release is the addition of Gemini and Anthropic as supported model providers, backed by a normalized authentication setup. You are no longer locked strictly into Qwen's own checkpoints to run the workflow.

For non-interactive and ACP usage, the update introduces /compress and /summary commands.

There is also a new experimental Skills system for extending capabilities, alongside handy editor tweaks like clickable bash toolcall outputs in VS Code. Under the hood, the team fixed flaky integration tests, smoothed out Windows CLI path resolution, updated the OAuth client for Figma's MCP server, and streamlined SDK releases.

Why it matters

Opening up to rival models like Claude and Gemini is a smart move. Developers rarely want to switch their entire CLI setup just to try another foundation model. By normalizing authentication across multiple providers, Qwen Code shifts its value from a simple model showcase to a genuine, agnostic developer tool.

Tackling unglamorous friction points like Windows path issues and flaky tests shows the team cares about daily reliability, which is usually what determines whether a terminal utility actually sticks in a developer's real-world stack.