Qwen Code 1.0 Is Here — and It's Not Just a CLI Anymore

Alibaba dropped Qwen Code 1.0 today, November 11, 2026. I read the announcement twice because the first sentence didn't sound like a version bump — it sounded like a change of job description.

Qwen Code used to be a terminal coding assistant. Now it's an open-source agent harness you can embed inside your own software. That's a different product wearing the same name.

What actually shipped

The short version: Alibaba merged Qoder's agent harness capabilities into Qwen Code and open-sourced the whole thing together as Qwen Code 1.0. Qoder was Alibaba's agent-assisted development tool; its harness — the machinery that plans, calls tools, and keeps an agent on track across a long task — now lives inside Qwen Code.

The architecture tells you where this is going. It starts from an in-process, in-memory event bus and evolves toward a distributed multi-node, multi-session design, with a unified event model that decouples reasoning orchestration from tool execution. In plain English: the part that thinks and the part that does are separate, and the events between them are a first-class thing you can observe, route, and replay.

That's the difference between a coding CLI and an agent runtime. A CLI runs a loop. A runtime lets you build your own loop on top of it.

One honest flag, up front: the official docs carry a "Qwen Code 1.0 Preview" note. The 1.0 Preview is a separate branch, and the stable branch is retained. So this is 1.0, but it's a preview branch — treat it as the direction of travel, not the finished road.

How to try it

The install is the same as before. Standalone script on Linux and macOS:

curl -fsSL https://qwen-code-assets.oss-cn-hangzhou.aliyuncs.com/installation/install-qwen-standalone.sh | bash

Or via npm, if you prefer:

npm install -g @qwen-code/qwen-code

Then cd your-project and run qwen. First launch asks you to connect a model provider. Qwen Code works with Qwen, OpenAI-compatible providers, Anthropic, Gemini, and local providers. There's also a VS Code extension in beta now, if you'd rather not live in the terminal.

It's Apache 2.0 licensed, which is the part that actually matters for what comes next.

The part nobody's talking about enough: the billing

Here's the catch, and it's worth stating plainly because it's the thing that will surprise people. Model usage is billed by the provider you connect. The free Qwen OAuth tier ended in April 2026. If you sign in with a Qwen Plus or Pro subscription, that unlocks Qwen Code and Agent access; otherwise you're paying per token through QwenCloud, ModelStudio, or whichever provider you wire up.

Qwen Code the software is free and open source. Running it is not free. That's a normal arrangement — Claude Code works the same way — but if you remember the early days when Qwen Code ran on a free OAuth tier, that door is closed, and has been for months.

What "embeddable agent runtime" actually means for you

This is the question I kept coming back to, because "agent harness" is one of those phrases that sounds impressive and means nothing until you put it next to a real decision.

If you just want a coding assistant in your terminal, nothing changes today. Install it, point it at a model, ask it to fix the bug. The 1.0 Preview doesn't make that workflow different.

It changes things if you're building agents yourself. Right now, most teams building an agent face an annoying choice: use a closed harness like Claude Code's internals (great, but you don't own it), or assemble your own from LangChain-style parts and spend a month on plumbing — event routing, tool execution, session state, retries.

Qwen Code 1.0 is a third option: a real, production-used harness, open source, with an event model you can hook into. You can embed it, observe its events, and build your own agent product on top of a runtime that Alibaba itself ships in Qoder.

That's genuinely interesting. It's also where I'd slow down. An embeddable runtime is only as good as its docs, its stability guarantees, and its community — and a preview branch has none of those settled yet. The event model is the promise; whether third-party developers can actually build on it without chasing breaking changes is the test, and it hasn't happened yet.

Should you care?

A quick decision checklist:

• You use a coding CLI daily and want the newest Qwen Code: install the preview branch, point it at your provider, and see if the harness feels different on long tasks. Keep the stable branch around — preview means preview.
• You're building your own agent product: this is the most interesting open-source agent runtime to land in a while. Read the event model docs before you commit to anything.
• You're happy with Claude Code or Cursor: nothing in this release forces a switch. The value here is for builders, not for people who just want a better autocomplete.
• You were hoping the free tier came back: it didn't. Budget for tokens.

The bigger picture

Alibaba shipped two things today: this, and the Qwen4-Max preview. Read together, they point the same way — Alibaba wants the Qwen stack to be the place where agents are built, not just the model behind someone else's agent. An open runtime plus a flagship model is a platform play, and it's a credible one.

But platforms are judged on what developers build with them, not on launch day. Qwen Code 1.0's real test starts now: does anyone embed it, and does it survive contact with real products?

I'll be trying the preview branch on a real repo this week. If the event model holds up on a long, messy task, I'll write the follow-up. If it falls over, I'll write that too.

Have you tried embedding an open-source agent harness in your own product — Qwen Code or anything else? What broke first: the tooling, the docs, or the model?

---

Sources: IT Home via PANews, November 11, 2026; official Qwen Code documentation (qwenlm.github.io/qwen-code-docs). Qwen Code 1.0 is a Preview branch; the stable branch is retained. Model usage is billed by the provider.