Providers & ecosystem · 1 min read · Updated 2026-10-05
Open-weight vs closed models
Closed models
You send your question to the company's computers. You get top quality and no setup. But you pay per use, depend on the company, and your data goes to them under their rules.
Open-weight models
You run the model on your own machines or a host you pick. You control your data and can customise it. But you need strong computers and some know-how.
Is open-weight the same as open source?
Not always. Most share the model files but not the data or full recipe, and some have limits on how you can use them. Read the licence.
Key takeaways
- Closed: easy and powerful, but you depend on the company.
- Open-weight: more control, more work.
- Always read the licence.
Quick questions
Is open-weight free?
The files are often free, but running them costs computer time.
Which keeps data more private?
Running it yourself keeps data with you. Closed services publish rules on how they handle data.
Can I customise closed models?
Some companies allow it for some models.
Keep learning
LLM providers: who builds the modelsA few companies build the best-known chatbots: OpenAI (ChatGPT), Anthropic (Claude), Google (Gemini), Meta (Llama)…1 min readRead →How LLMs are trainedChatbots learn in three steps. First they read a huge amount of text to learn language. Then they practise on good example…1 min readRead →