UbikUbik
← All posts

Model Control Is Great, but Training Control Is Better

Choose models in Ubik and understand why training control matters. Recommended models, privacy, and when to use different model types.

Ubik Team
Model Control Is Great, but Training Control Is Better

Ubik gives you control over which models you use. Often more important than raw power is whether a model trains on your data. Ubik's recommended models do not train on your chat sessions, which matters for sensitive work.

Models in Ubik

Open the model toggle in Settings or the chat UI to choose which model handles your tasks.

Training Control and Recommended Models

Training control means knowing whether a provider uses your data for training. Ubik's recommended models are chosen so they do not train on your chat sessions. That matters when working with:

  • Confidential documents
  • Client data
  • Sensitive research
  • Personal or proprietary information

Model control vs. training control:

  • Model control: Which model runs your task (GPT-4, Claude, local, etc.)
  • Training control: Whether your data is used to train future models

Both matter, but training control is often the deciding factor for sensitive work.

Model Routing

You can route tasks to:

  • Frontier cloud models (GPT-4, Claude, Gemini)
  • Experimental models
  • Local models for privacy or offline use

You choose what runs where and when. Multi-model routing can automatically pick a model for each task, or you can set it manually.