Prompt template model selection
Key Takeaway
To stop people from selecting a model in Prompt Builder while existing prompt templates keep running, hide its LLM configuration in AI Models. Turning off the model provider blocks all of its models and stops affected components from running.
In Prompt Builder, every prompt template runs on a large language model (LLM) configuration that you choose. The right model is the one that tests best on your own task and data, not the newest one. Admins also control which models people can use. They can hide one LLM configuration from selection in Prompt Builder, or they can turn off a whole model provider for the org. The two controls affect prompt templates that already use the model very differently.
Where a prompt template's model comes from
You can change a prompt template's LLM configuration in Prompt Builder with a few clicks.
Standard configurations come included with Prompt Builder. Custom configurations are brought in through BYO-LLM. If you modify a standard configuration, the changed configuration is saved as a custom configuration.