In my past posts, one of the user from Hive asked me to sooner or later switch to the CLI. And turns out the suggestion has worked out. The way CLI has improved my usage pattern and also the habit has helped me with the productivity. Like it's stop you from remembering too much stuff. Just a prompt and process to get the product done.
Before that let me tell you that I used Warp command prompt. And it comes with 150 CLI queries to the AI models. And turns out those models are firing pretty rapidly for the Claude code. And turns out the models and the output is working out. In that context I would say that Claude models beat all of the other code models.
So ignoring that part let's move onto our experience with two models : QWEN and Gemini.
I downloaded Gemini CLI first, it seems there is same interface for both QWEN and the Gemini CLI.
Gemini CLI has open sourced their CLI interface, so I suppose QWEN picked it up and used it effectively. Token wise both the models are offering better pricing and also the free usage is also decent for the both models. However the experience with the prompts vary.
For example, I created Termina using the Warps claude model. While with the QWEN I easily turned the markdown to html5 and it was a good update. Instantly it worked out less than few minutes. Whereas the same prompt given to the Gemini and it kind of yet to give me output and is stuck.
https://x.com/devnamipress/status/1959283473849430074
This made me stick with the QWEN instead of Gemini for certain such operations. I prefer making use QWEN model now because it has shown more results to me than the Gemini itself. Which goes to show that the models like these make better output and are suitable for the vibe coding or any form of coding at all.
To each their own. some people have a good experience with the Gemini whereas some people would be finding claude and qwen to be pretty effective. Use what works for you.