Collab_GPU

Use Google Collab for free GPU access. Use Google free GPU for updating your code via OpenCode or Claude

---

Install and run xterm with following commands:

1. Goto Google Colab and Create a New Notebook.
2. In the upper right corner -> Click down arrow and change runtime type to "TPU" from "CPU". Click save.
3. Connect.

Once GPU is connected; we need to install XTERM. This is used to run ollama in the background (and not on our main notebook interface). Use below 3 commands to do that.

!pip install colab-xterm

%load_ext colabxterm

%xterm

Once this is done you will see XTERM started as below screenshot

In the new Terminal Run Following Commands (not in notebook/ in terminal)

apt install screen apt install zstd

How to go inside screen and get out of it

To go inside screen type: screen To get outside type: CTRL A + CTRL D Reconnect to same screen again: screen -r (reconnect)

Inside screen, run following commands:

curl -fsSL https://ollama.com/install.sh | sh ollama serve

Outside screen, on main terminal, run following command:

// Since the GPU RAM is 15GB useable; I will choose a model almost half a billion in parameters to run smoother. I choose gemma4:latest. You can also pull ollama pull qwen3.5:9b ollama run qwen3.5:9b --verbose

Now I will install opencode (you can also install claude/ gemini) and connect it with gemma4 on Google Free TPU