Choice of AI model
Being locked into one AI provider is a risk. Prices change, models are retired, a provider has a bad day, or your own company only approves certain vendors. ChatHQ is built so that the model behind an assistant is a setting, not a foundation.
Picking a model
When you create or edit an assistant, you choose its model from the model catalogue available to your workspace. The catalogue is grouped by provider, for example OpenAI, Anthropic and Google, and each entry shows the model name and its status.
Which models are available can vary by workspace and plan.
Why it matters
- Your policies, not ours — If your organisation approves only certain providers or models, pick from those. Nothing else about the assistant changes.
- The right model for the job — Some models answer faster, some reason better over long documents, and some cost considerably less. You can choose per assistant: a cheaper model for a simple FAQ bot, a stronger one for complex policy questions.
- Resilience — If a provider has problems, switching an assistant to another model is a configuration change, not a migration.
Changing the model
Changing the model edits the assistant's draft, like any other setting. Instructions, knowledge sources and widget settings are untouched. Test the draft in the test chat, then publish. Visitors keep talking to the current live version until you do. See Assistant versions.
Different models phrase things differently. After switching, run your most important questions through the test chat and adjust the instructions if the tone or format has shifted. See Instructions & behavior.
Related
Was this article helpful?