Selecting the models to use

From the Data Agent chat, you can select the specific models to use for different types of tasks, based on the model performance and model cost. For example, for a larger data generation job, you might select a lower-cost model.

You also select the level of effort and the maximum number of tokens for the chat context, both of which also affect the performance and cost.

Your account owner can change both the default and available models for the account.

To change the selected model and level of effort:

1. Click the current data generation model and effort selection.

Current model and effort selection for the Data Agent

1. On the Models panel, under Model:

   1. Under Data Generation, click the radio button for the model to use to generate date.
   2. If validation is enabled, then under Validation, click the radio button for the model to use for validation.
   3. Under Profiling, click the radio button for the model to use for data profiling.

   The default and recommended option for all three is Sonnet 5.5. Note that if you use your own LLM provider, and are not connected to Azure AI Foundry, then the GPT models are not available.

Models panel with the model and effort selection

1. Under Effort, click the radio button for the level of effort to use. The default is High.
2. Under Freeform text model, click the radio button for the category of model to use to generate freeform text. Fabricate uses the model that it currently has mapped to that category.

   • Economy - Lowest quality and cost. Medium speed.
   • Fast - This is the default and recommended option. This is the fastest option, with high quality and medium cost.
   • Standard - Provides higher quality at a higher cost, but with slower speed.
   • Premium - Provides the highest quality and the highest cost, but is very slow.
3. From the Context Window dropdown list, select the maximum number of tokens for the chat context. The default is 200,000 tokens.
4. Click Done.