Gemma 4 is Good 🤖

Last month, I talked about Bolt.diy where I used Qwen2.5-code to make local AI vibe coding. 💻

There is a website, canirun, where I can find the models that I can run on my hardware. 🔍

Recently, my YouTube timeline was booming with news about how good Gemma 4 is. 📈 I thought it was just hype at first. But as I changed to Linux, I have more resources to access. I decided to try it. 🐧

I downloaded gemma-4-e4b. It has 7.5B parameters (4 experts) and is only 6.3GB. 📦 I changed the load into max, context length to 131,072, and GPU Offload to 42. These parameters allow it to perform tool-use, such as creating files. 🛠️

Then, in VS Code, I installed the Continue extension. It can connect to a local model. Change the config.yaml as follows: ⚙️

name: Local Config
version: 1.0.0
schema: v1
models:
  - name: LM Studio
    provider: openai
    apiBase: http://localhost:1234/v1
    model: lmstudio
    roles:
      - chat
      - edit