I’ve been using Cursor for a while, mainly for its smooth AI coding experience. But recently, I decided to move my workflow back to VS Code and test how far open-source coding models have come.
The setup I’m using is simple:
- VS Code + Hugging Face Copilot Chat extension
- Models: Qwen 3, GLM 4.6, and Kimi K2 from Nebius AI Studio
Honestly, I didn’t expect much at first, but the results have been surprisingly solid.
Here’s what stood out:
- These open models handle refactoring, commenting, and quick edits really well.
- They’re way cheaper than proprietary models, no token anxiety, no credit drain.
- You can switch models on the fly, depending on task complexity.
- No vendor lock-in, full transparency, and control inside your editor.
I still agree that Claude 4.5 or GPT-5 outperform in deep reasoning and complex tasks, but for 50–60% of everyday work, writing code, debugging, or doc generation, these open models perform just fine.
It feels like the first time open LLMs can actually compete with closed ones in real-world dev workflows.
I also made a short tutorial showing how to set it up step-by-step if you want to try it:
I would love to hear your thoughts on these open source models!
Top comments (0)