BeginnerStart here. No prior knowledge assumed.
0 of 5 complete
Lesson 1
Open vs Closed Models
What open weights actually means, and why it is not the same as open source.
Lesson 2
Llama
Meta's open-weight family and the licence terms that come with it.
Lesson 3
Mistral
European open-weight models with permissive licensing.
Lesson 4
DeepSeek
Near-frontier reasoning at open-weight cost, and the data residency question.
Lesson 5
Qwen
Alibaba's wide size range and strong multilingual performance.
What does open weights mean?
What is the first thing to check before commercial use?
IntermediateBuild on the basics.
0 of 5 complete
Lesson 1
Running AI Locally
Ollama, quantisation and what your hardware can genuinely handle.
Lesson 2
Small Language Models
Distillation, the cascade pattern, and where small beats large.
Lesson 3
Kimi
Long-context open models and the lost-in-the-middle problem.
Lesson 4
Token Economics
Why self-hosting changes the cost calculation entirely.
Lesson 5
AI Evals and Benchmarks
Building your own eval set, because public benchmarks will not decide this.
What does quantisation do?
What is the cascade pattern?
AdvancedDepth, edge cases and strategy.
0 of 5 complete
Lesson 1
Fine-Tuning
Adapting an open model to your domain, and when it beats RAG.
Lesson 2
RAG
Giving a self-hosted model access to your own documents.
Lesson 3
Embeddings and Vector Databases
The retrieval layer you will also need to self-host.
Lesson 4
Reasoning Models
Test-time compute in open models, and when the extra cost pays.
Lesson 5
AI Regulation
Obligations that apply regardless of whether the model is open.
Does self-hosting resolve the data residency concern?
What should decide between open models?