How to use local AI in Coditan

Before even thinking of downloading a model you must first know what AI you'll use
You can use the table grid below to find the best match for you :

Note
All of the VRAM usage is calculated with the Q4_K_M quantization used by almost all models uploaded to Ollama.
This list is a near estimate of what we expect with about a 32k context window using specific Optimization Ollama has built in

GPU VRAM Model Size
Nvidia RTX 3050 6gb 3 billion parameters
Nvidia RTX 3080 12gb 7 billion parameters
Nvidia RTX 4070 TI Super 16gb 14 billion parameters
Nvidia RTX A4500 20gb 24 billion parameters
Nvidia RTX A5000 24gb 35 billion parameters
Nvidia RTX A6000 48gb 72 billion parameters

How to install & use a model

For this guide We will use Ollama as it provides a massive library of easy to run models and can easily be connected

  • 1. Download Ollama: To download ollama you'll need to go to The Official Ollama downloads
  • 2. Download a Model: After installing ollama you need to install a model, it might be confusing so we want to help you by giving a list of the best models to run!
3 billion parameters 
ollama pull qwen2.5-coder:3b
7 billion parameters 
ollama pull qwen2.5-coder:7b
14 billion parameters 
ollama pull qwen3:14b
24 billion parameters 
ollama pull devstral-small-2:24b
35 billion parameters 
ollama pull qwen3.6:35b
72 billion parameters 
ollama pull qwen2.5vl:72b

After installing the model you'll need to open Coditan and follow these simple steps :

  • Coditan > Settings (Right top) > Preferences > Model
  • After that simply paste like this
ollama/the-model-you-choose
  • Example of the correct usage
ollama/qwen3.6:35b

Tip
The way you write the model is very simple
Ollama - The Provider
Qwen3.6:35b - The model you want to use
So it's just provider + model with a /in the middle.