What is the underlying LLM?

Short Course Q&A Build with Andrew!
1

Want to learn to tweak the LLM too along with building the project. Currently we can only see the resulting HTML file- but could you please explain what is the underlying LLM and if it is opensource, how can we use it in our projects? (GPT and Claude are paid, but maybe this is Ollama? can’t find a good course on ollama either- that is a separate issue though)

2

Hi @Deepashree_Sengupta

Thanks for learning with us! Let me share some answers:

1.“What LLM is actually powering this?”
If you only see the resulting HTML file, the LLM itself isn’t in that file — it’s whatever generated the code (a course platform’s backend, an AI website builder like Bolt/v0/Lovable, or a custom app calling an API). Good habit to explore more: open browser dev tools → Network tab → look for calls to api.openai.com, api.anthropic.com, or a custom backend URL. That’s usually the fastest way to reverse-engineer “what’s under the hood” of any AI tool, not just this course’s project.

2. “Is it open source?”
Yes, and there’s a real ecosystem for it now:

  • Ollama is the tool, not a model. It’s a free runtime that downloads and runs open-weight models locally (no API key, no per-token cost).
  • The actual open models people run through it include Llama, Qwen3, DeepSeek, Gemma, and OpenAI’s own open-weight release (gpt-oss) — Apache 2.0 licensed, run entirely on your own hardware with no API key required.
  • Ollama installs as a single binary, and you pull a model with ollama pull llama3 then chat with ollama run llama3 or use its OpenAI-compatible API on port 11434 — which means most existing OpenAI/Claude-style code works by just swapping the base URL.
  • For a first model on a laptop, an all-rounder pick is qwen3:30b if the machine can handle it, or qwen3:14b if not.

3. “Can’t find a good Ollama course.”
Fair — most content out there is scattered blog posts, not structured courses. Practical fix: check out this 20-minute checklist:

  1. Install Ollama from ollama.com
  2. ollama run llama3.2 (or qwen3) → confirms it works
  3. Point their existing project’s API calls at http://localhost:11434/v1 instead of OpenAI/Anthropic
  4. Compare outputs — this is “tweaking the LLM”: swapping models, adjusting system prompts, testing different sizes.

Hope this helps! Happy learning.

-- Lesly, DLAI