Aug 15, 2026 · 4 min read

I Used Qoder for a Week — Alibaba's Answer to Claude Code

This is week 20 of my “I Used It for a Week” series. Last week I reviewed ZCode, the desktop agent with Telegram remote control. This week: the tool that emerged from Alibaba’s Claude Code ban.

Qoder is what happens when a company bans Claude Code and needs a replacement fast. Alibaba built Qoder as an internal tool, then made it available externally when the ban became public.

After a week of using Qoder on real projects, I think it’s a solid CLI agent that’s held back by its rushed origin. It works, but it feels like it was built in a hurry.

How It Works

Qoder is a terminal-based coding agent, similar to Claude Code. You install it, point it at your project, and start talking.

npm install -g @alibaba/qoder
cd my-project
qoder

It reads your files, makes changes, and runs commands. The interface is familiar if you’ve used Claude Code or Aider.

Day 1: First Impressions

I opened a Python project and asked: “Add async support to all synchronous HTTP calls.”

Qoder identified the synchronous calls, refactored them to use aiohttp, updated the imports, and ran the tests. Total time: about 5 minutes.

The code quality was decent but not great. It used aiohttp instead of httpx, which is more modern. And it didn’t handle some edge cases that Claude Code would have caught.

Day 2-3: The Chinese Model Integration

Qoder is built on Qwen, Alibaba’s flagship model family. The coding capabilities are solid — it handles Python, JavaScript, TypeScript, Go, and Java well.

The interesting part: Qoder has built-in support for Qwen’s thinking mode. For complex tasks, you can see the model reason through the problem before producing code. The thinking is useful but sometimes too verbose.

I tested it on a complex refactoring task: migrating from Express to FastAPI. Qoder handled it, but needed more guidance than Claude Code. It missed some edge cases and I had to point it in the right direction a few times.

Day 4-5: Speed and Cost

Qoder is fast. Qwen’s inference is optimized for speed, and it shows. Simple tasks complete in seconds. Complex refactoring is noticeably faster than Claude Code.

Cost-wise, Qoder is very affordable. Alibaba’s pricing is aggressive, and the free tier is generous. For cost-conscious developers, this is a major advantage.

The Telegram integration (similar to ZCode) lets you monitor tasks from your phone. It’s less polished than ZCode’s implementation but functional.

What Blew Me Away

Speed

Qwen’s inference is genuinely fast. For tasks where speed matters, Qoder beats Claude Code and MiMo Code.

Cost

The pricing is aggressive. Free tier is generous, paid tiers are affordable. For high-volume coding, this matters.

Multi-language support

Qoder handles Chinese programming documentation and comments natively. If you work with Chinese developers or Chinese codebases, this is a unique advantage.

What Frustrated Me

Code quality

The code Qoder generates is functional but not as polished as Claude Code’s. It makes more mistakes, misses edge cases, and sometimes uses older patterns.

Documentation

The documentation is in Chinese and English, but neither is comprehensive. Figuring out advanced features took trial and error.

No git integration

Unlike Aider, Qoder doesn’t automatically commit changes. You need to review diffs and commit manually. This breaks the git-native workflow that Aider pioneered.

Limited ecosystem

Qoder is new and the community is small. Finding help, tutorials, or integrations is harder than with Claude Code or Cursor.

Language support

Qoder excels at Chinese programming contexts. If you’re working with Chinese developers, Chinese documentation, or Chinese codebases, Qoder handles this natively. English-only developers may not notice this advantage.

Speed comparison

Qoder consistently outperforms Claude Code and MiMo Code on response speed. For tasks where latency matters, Qoder is the fastest option I’ve tested.

Qoder vs Claude Code vs Aider

Claude Code: Best model quality, best ecosystem. But costs more and is US-focused.
Aider: Best git integration, cheapest, most flexible. But no visual UI.
Qoder: Best speed, best Chinese language support. But less polished code quality.

Would I Keep Paying?

Qoder’s free tier is generous enough for most use cases. The paid tiers are affordable. If you need speed and cost efficiency, Qoder is worth considering.

Rating: 7/10 — Fast, cheap, and functional. But the code quality and documentation need work.

When to Choose Qoder Over Alternatives

You work with Chinese codebases. Qoder’s native Chinese support makes it the best choice for Chinese projects.

Speed is your priority. Qoder is the fastest CLI agent I’ve tested. If latency matters, Qoder wins.

Budget is tight. The free tier is generous and paid tiers are affordable.

You need Alibaba ecosystem integration. Qoder works well with Alibaba Cloud services, Qwen models, and Chinese developer tools.

You’re okay with less polish. Qoder works but lacks the refinement of Claude Code. If you’re comfortable with a rougher tool, the speed and cost savings are worth it.

FAQ

What is Qoder?

Qoder is Alibaba’s terminal-based coding agent. It emerged as an internal tool after Alibaba banned Claude Code, then was made available externally.

How does Qoder compare to Claude Code?

Qoder is faster and cheaper but less polished. Claude Code generates higher-quality code and has better documentation. Qoder wins on speed and cost.

Is Qoder free?

Qoder has a generous free tier. Paid tiers are affordable and competitive with other coding tools.

Does Qoder support thinking mode?

Yes. Qoder has built-in support for Qwen’s thinking mode, which lets the model reason through complex problems before producing code.

Next week: I Used Codex CLI for a Week — OpenAI’s terminal coding agent.