Open WebUI and Ollama Notes

Browse available models at the Ollama model library.

---

Model Recommendations

General purpose:

• Mistral -- Fast and efficient, great for a mix of tasks
• LLaMA -- Solid for creative writing, reasoning, and general knowledge
• Gemma -- Designed with responsible AI principles in mind

Compact and lightweight:

• Phi-2 -- Small but surprisingly capable, good for quick responses
• TinyLLaMA -- Optimized for speed and minimal resource usage

More depth in responses:

• LLaMA 2 13B or 70B -- Larger models with better reasoning
• Mixtral -- Mixture-of-experts model with boosted performance

Running on a home server: bigger models need more VRAM and compute. For a fast, lightweight setup Mistral or Phi-2 are good starting points.

---

Working with Models

List installed models:

ollama list

Pull a model from the Ollama repository:

ollama pull mistral

Replace mistral with any model name from the library (e.g. llama2, gemma, phi).

Set the model in Open WebUI:

Once pulled, select the model in Open WebUI settings or when starting a new conversation.