I’ve been using Qwen 3.6 Plus since last night, trying to decide if the new model is worth it, after part of the team (including the tech lead) left Alibaba. To be honest, 3.5 did not live up to the expectations (at least mine, and many others’).
What I’m using it for:
Well, to be honest I’ve been vibe-coding an app to make my life easier. I’m in the progress of migrating from Windows to Linux. On Windows I’m using Laragon for my web-dev, as it ticks most of my boxes:
- Low ram usage
- Multiple PHP versions, switching on the fly
- Custom hostnames (I hate using https://localhost/projectname, breaks compatibility with badly written sites)
- Mail catcher (so you don’t have to wonder if your mail-sending script works)
and many more. However, it is windows-only.
I’ve been using Minimax M2.7 to build an alternative for Linux, that uses caddy as a webserver, multiple PHP versions (with planned feature to use seperate PHP versions per project), etc.
The stack is go with wails & vite. I’m planning to switch to a Macbook in the future, and everything is implemented to use apt or brew (even though I can’t really test it in OS X right now)
Why I’m switching
After the codebase grew a little bit, Minimax was struggling a little bit to implement new features. It was taking too long to locate the correct parts of where to add the generated code (even though I was providing the files where the code was supposed to go), and was making mistakes that it was fixing earlier.
I had to ask it to fix an issue 3 to 4 times, but it got to the point where it wasn’t working out for me.
Then I noticed that I had access to the newly announced Qwen 3.6 Plus model through Opencode Zen & Openrouter.
How’s it going so far?
It is really fast in fixing things, and implementing things. The 1M token window is really great. No hallucinations so far (I had that issue with 3.5), no super-extensive reasoning/thinking sessions.
It may not be great for everyone, but it is working great for me. I just hope they won’t make it dumber when it is released, as it is a “preview” right now.
If you don’t mind possible data-training (I don’t for this specific project, as it is fully generated through LLMs), give it a go.
My only hope is that Alibaba reinstates the $10/month Coding plan, as the $50/month plan is right now to expensive for what it offers.