Providers and models¶
Last reviewed: 2026-08-14
A model provider is a company or service that runs a model and lets your application call it through an API.
I use LiteLLM in the Course because each provider expects slightly different request code. LiteLLM is a Python library that gives us one common function while still letting you choose the provider and model.
Provider prices, free limits, privacy terms, and model names change. Check the linked provider pages before you deploy anything.
Options for following the Course¶
| What you need | Option | What to know |
|---|---|---|
| Free answer generation | Google Gemini free tier | Free request limits vary. Review how free-tier data may be used. |
| Inexpensive paid generation | OpenAI GPT-5.6 Luna | Requires billing and gives more repeatable capacity than a free tier. |
| Fast experiments | Groq | Free request limits and available models can change. |
| Compare several providers | OpenRouter | One API gives access to many models, but free capacity is not a reliability promise. |
| Create embeddings on your computer | sentence-transformers/all-MiniLM-L6-v2 |
Runs without an API key and works on a normal processor, but you still need to evaluate it on your own Sources. |
Hugging Face Inference Providers gives free accounts a small monthly credit. I do not present it as a large free generation allowance.
Keep API keys in environment variables instead of source code. Record the exact LiteLLM and model versions with every evaluation run so you can reproduce a result later.