Learn

Answers to the questions people ask before running a model themselves.

• How do I run a large language model on my own hardware?
• When should I use a LoRA adapter instead of fine-tuning a whole model?
• How do I use an AI coding assistant without sending my code to a vendor?
• How can one GPU serve many specialised models at once?
• Is a self-hosted AI assistant actually cheaper than paying per seat?
• What is model routing, and why does an AI orchestrator need it?
• Can I train a domain specialist model without an ML engineering team?
• Which quantisation should I choose to run an LLM locally?
• How much VRAM do I actually need to run a 70B model?
• Which open-source models can realistically replace ChatGPT for my work?
• How do I fine-tune a model on my own documents without leaking them?
• Do I need retrieval-augmented generation or fine-tuning?
• Can an AI assistant work with no internet connection at all?
• Which GPU should I buy to run AI models at home?
• Is Apple Silicon or an NVIDIA GPU better for running LLMs locally?
• How do I stop a vendor from training on my company's data?
• How do I serve a language model to my whole team from one server?
• How do I tell whether my fine-tuned model is actually better?
• What does it really cost per month to run your own AI server?
• What is a LoRA adapter, in practical terms?