CodeGPT setup: Ollama Llama 3 8B vs Groq Llama 3 70B

This tutorial, uploaded May 30, 2024, shows how the speaker configured CodeGPT as a coding assistant in VS Code. It uses Llama 3 8B through Ollama and Llama 3 70B through Groq. The source gives no CodeGPT or Ollama release number, so the interface and setup reflect the upload's historical context.

The local setup starts with installing the extension from the VS Code marketplace. The speaker then downloads Ollama, pulls the 8B model and checks it in the terminal before selecting Ollama as CodeGPT's provider. A selected class becomes the first test: the extension sends the code to the local LLM and returns an explanation inside the editor, avoiding the speaker's usual copy-and-paste workflow with ChatGPT.

The second setup uses Groq's hosted inference service. The speaker creates an API key, changes the provider in CodeGPT and selects the 70B model. This route differs from using Ollama to run models locally: the demonstration connects to Groq through an API.

The speaker finds the 70B response more detailed, particularly when explaining regular expressions, and demonstrates inserting a suggested refactor into the file. The generated indentation still needs attention. These examples support a comparison of the responses shown, rather than a general benchmark of either model. The video's free-use framing belongs to its 2024 context and does not establish current service terms.