Models For Coding
After using local LLMs for TypeScript and JavaScript development for over a year, I’ve learned that model choice makes a huge difference. Not all models are created equal when it comes to coding, and some are terrible at certain languages while being great at others.
Here’s what I’ve discovered about which models actually work well for TypeScript/JavaScript development, and more importantly, why.
My Experience with Coding Models
I primarily work with TypeScript for backend services and React frontends, with some vanilla JavaScript thrown in. I’ve tested dozens of models specifically for these use cases, and the differences are dramatic.
What I’ve learned:
- Model size isn’t everything - a well-trained 7B model often beats a poorly trained 13B model
- Some models are great at explaining code but terrible at generating it
- TypeScript support varies wildly between models
- Most models trained primarily on Python struggle with modern JavaScript patterns
What This Section Covers
Model Comparisons: Which models I actually use and why Licensing Guide: What you can legally do with each model Fine-tuning Notes: My experiments with customizing models for TypeScript work
This is purely about model selection - I’m not covering implementation details or tutorials here.