Configure your model, generate a ready-to-run Python script, paste it into Colab, and get a public chat URL in minutes.
Running the script with !python3 file.py or pasting directly into a Colab
cell will NOT work properly โ Cloudflare URLs, rich progress bars, and
interactive CLI chat require a real terminal, not a notebook cell.
Instead, upload the file and run it from the Colab terminal as described below. ๐
Click the ๐พ Save as .py button above to download the generated script
as a .py file to your computer. Or use ๐ Copy Code and
save it manually.
Go to colab.research.google.com and create a new notebook.
Click Runtime โ Change runtime type and select T4 GPU (free
tier) or A100 (Colab Pro). Larger models need more VRAM.
In the left sidebar, click the ๐ Files tab, then click the โฌ
Upload button. Upload the .py file you saved in Step 1. It
will appear in /content/.
Click Runtime โ Run in Terminal, or use the keyboard shortcut. You can
also type !bash in a cell and press Enter to open a shell. This gives you a
real terminal that shows all output correctly.
In the terminal, run:python3 /content/your_model_colab.py
Replace the
filename with whatever you saved it as. The script will install everything, download the
model, and show progress bars.
A Cloudflare tunnel will create a public URL like
https://xyz.trycloudflare.com. This URL will be clearly displayed in the
terminal output. Share it with anyone!
Use the web chat UI via the public URL, or use the CLI
chat directly in the terminal. Type exit to quit CLI
mode.
โข Always use the terminal โ never !python3 in a cell. The
! prefix breaks rich output and Cloudflare URL detection.
โข Free Colab gives you ~12.7 GB VRAM with T4. Models up to ~14B parameters (quantized) fit here.
โข Session time limit: Free Colab disconnects after ~90 minutes of idle time or ~12 hours total.
โข Don't close the tab while the script is running โ the Colab session needs to stay active.
โข Cloudflare URL changes each time you restart โ share the new one.
โข For models > 14B parameters, use Colab Pro with A100 GPU (40/80 GB VRAM).