Ollama
Open sourceRun open models on your machine. The privacy and outage plan for Myanmar builders.
No cardMM-friendlySelf-hostableAPI
Free tier
Free local runtime. Unlimited once the model is downloaded.
Limits
Quality trails frontier APIs. Needs RAM/GPU. First download is heavy.
Best for
- Offline work
- Private code
- Zero marginal cost
How to use it well
- Install Ollama, pull a small model first (`qwen2.5:7b` or current Gemma).
- Point Cline/Aider/OpenCode at `http://localhost:11434`.
- Upgrade model size only after you know RAM headroom.
Myanmar notes
Download models on Wi-Fi. A 7B model is the realistic laptop target in MM.