Qwen Code Weekly: Channels Multi-Platform Access, Cron Scheduled Tasks, /plan Planning Mode, Qwen 3.6 Plus Launch

This week we released two feature versions v0.14.0 and v0.14.2, along with two bugfix versions v0.13.2 and v0.14.1.

v0.14.0 is a major update:

• The Channels system breaks Qwen Code free from the terminal, enabling remote operation through Telegram, WeChat, DingTalk, and other platforms;
• Cron scheduled tasks let AI automatically execute repetitive work on a schedule;
• Qwen 3.6 Plus model is officially launched. v0.14.1 brings follow-up suggestions, verbosity mode toggle, and other experience improvements.
• v0.14.2 adds the /plan planning mode command, thinking block cross-turn retention, adaptive output token upgrade (default 8K + auto-upgrade to 64K), and built-in bugfix workflow and debugging skills.

✨ New Features

Channels: Operate Qwen Code Remotely via DingTalk, Telegram, WeChat

Qwen Code is no longer limited to the terminal. The Channels system supports plugin-based access to multiple platforms—currently supporting Telegram, WeChat, and DingTalk. You can send a message to the Bot from your phone and have it execute tasks on your server.

What You Can Do With It:

• While out and about, send a message to Qwen Code via Telegram to run a script or check logs
• @mention the bot in a DingTalk group to handle code-related tasks for the team
• Send a quick message on WeChat to remotely operate your development environment

See PR #2628 

Cron Scheduled Tasks: Let AI Work on a Schedule

A new Cron tool lets you set up scheduled recurring tasks within the current session. AI will automatically execute them on your defined schedule without you having to watch.

What You Can Do With It:

• Automatically check if tests pass every 30 minutes
• Automatically pull the latest code and run a build every morning
• Monitor log files on a schedule and notify you when anomalies are detected

See PR #2731 

How to Enable?

Add the following to ~/.qwen/settings.json:

{
	"experimental":{
		"cron": true
	}
}

Qwen 3.6 Plus Model Launch

Qwen 3.6 Plus is officially available in Qwen Code, free to use. Its coding performance is on par with GPT-5 and Claude 3.7 Sonnet, with notable advantages in Chinese understanding and long-context processing. Alibaba Cloud ModelStudio Coding Plan is also now available.

What You Can Do With It:

• Use the latest Qwen 3.6 Plus model directly in Qwen Code
• Free 1,000 calls/day, 1 million token context
• Better experience for Chinese programming scenarios

See PR #2820 

Follow-up Suggestions: AI Tells You What to Do Next

After completing a task, AI automatically provides 2-3 follow-up suggestions. No need to think about “what should I do next”—just click to continue.

What You Can Do With It:

• After writing a component, AI suggests “Want to add unit tests?”—just click to start
• After fixing a bug, AI suggests “Want to check similar places?”—saves you the thinking
• When beginners don’t know what to do next, follow-up suggestions serve as navigation

See PR #2525 

How to Enable?

Add the following to ~/.qwen/settings.json:

{
	"ui":{
		"enableFollowupSuggestions": true
	}
}

For faster suggestion response times, you can use /model --fast to specify a lightweight model dedicated to background suggestions.

Sub-Agent Cross-Model Selection: Different Models for Different Tasks

Sub-agents can now use different models from the main agent. Use a powerful model for the main task to ensure quality, and a lightweight model for sub-tasks to speed things up—flexible combinations that save tokens without sacrificing quality.

What You Can Do With It:

• Use Qwen 3.6 Plus for the main task and a lightweight model for sub-tasks, reducing token consumption while maintaining quality for critical tasks
• Different types of sub-tasks are automatically assigned the most suitable model

See PR #2698 

How to Get Started?

Create a sub-agent Skill file specifying a different model:

---
name: test-cross-provider
description: Test cross-provider Agent
model: openai:gpt-4o
---
 
You are a test assistant. Please briefly introduce yourself, including the model name you are using.

Then tell Qwen Code to invoke this sub-agent in the conversation.

Ctrl+O Verbosity Toggle: Switch Output Detail Level with One Key

Press Ctrl+O to toggle between verbose (detailed) and compact (concise) modes. View detailed output when debugging, switch to compact mode for daily use—no configuration changes needed.

What You Can Do With It:

• Switch to verbose mode when debugging to see the full execution process
• Switch to compact mode for daily use for a cleaner interface
• Toggle anytime without restarting or editing config files

See PR #2770 

/plan Command: Plan Before You Act

The new /plan command lets you enter planning mode with one command. AI will first provide a complete execution plan, and you confirm before it starts executing. Perfect for complex tasks—see the full picture first, then proceed step by step.

What You Can Do With It:

• Before refactoring a large module, have AI list all files and steps that need to be changed
• Review the plan for multi-step tasks first to avoid AI going off track and having to start over
• In team collaboration, generate a plan for colleagues to confirm before executing

See PR #2921 

Adaptive Output Tokens: Responses No Longer Get Truncated

Output tokens default to 8K. When AI detects a response is being truncated, it automatically upgrades to 64K and retries. No need to manually adjust parameters—AI determines how much response space it needs on its own.

What You Can Do With It:

• Long file generation is no longer truncated—complete content in one output
• Complex code generation without needing to piece together segments
• No need to manually configure max_tokens—AI adapts automatically

See PR #2898 

Thinking Block Cross-Turn Retention

AI’s thinking process (thinking block) can now be retained across conversation turns, with automatic cleanup during idle time. In multi-turn conversations, AI can better continue its previous reasoning chain for more coherent responses.

See PR #2897 

📊 Improvements

🔧 Important Fixes

🎈 Other Improvements

👋 Welcome New Contributors

• @chinesepowered — Fixed Hooks abort listener cleanup, sub-agent cache refresh, Telegram message fallback, and more
• @pic4xiu — Fixed Hook system message commit timing, removed WebFetchTool duplicate proxy settings
• @YingchaoX — Restored ? shortcut in vim normal mode
• @euxaristia — Removed quote drag detection to fix input delay
• @kulikrch — /theme restores previous theme on cancel
• @nsalvacao — Fixed exit_plan_mode in YOLO mode
• @mj4444ru — Fixed Markdown table cell separator escaping
• @chiga0 — Implemented Ctrl+O verbosity mode toggle

If you have questions or suggestions, feel free to provide feedback on GitHub Issues !