Skip to main content
Your Chat Agent’s intelligence comes from a large language model (LLM). On the Basic tab, you can select which AI provider and model your agent uses. The model you choose affects response quality, speed, and capabilities. AI Provider and Model dropdowns on the Basic settings tab

​
Provider and Model Fields

FieldDescription
AI ProviderThe company providing the language model. Currently set to OpenAI for all agents.
ModelThe specific model version used to generate responses. Different models offer different tradeoffs in quality, speed, and cost.

​
Available Models

ModelDescription
GPT-4oA multimodal model optimized for speed and efficiency. Good balance of quality and response time.
GPT-4.1An improved version of GPT-4 with better instruction following, longer context handling, and more accurate responses. Recommended for most use cases.
GPT-5The latest and most capable model. Best for complex conversations that require advanced reasoning.

​
How to Select a Model

  1. Navigate to Agents > Chat Agents in the sidebar and select your agent.
  2. Click the Basic tab.
  3. Click Start Editing if not already in edit mode.
  4. Locate the AI Provider and Model dropdowns in the left column.
  5. Click the Model dropdown and select your preferred model.
  6. Click Save Changes.

​
How the Model Affects Your Agent

The model determines how well your agent:
  • Follows instructions — More advanced models adhere more closely to the objectives and role you define.
  • Handles complex questions — Newer models are better at multi-step reasoning, understanding nuance, and providing accurate answers.
  • Responds quickly — Lighter models like GPT-4o tend to respond faster, while more powerful models may take slightly longer.

​
Tips

  • Start with GPT-4.1 for most agents. It offers the best balance of quality, speed, and cost for typical customer-facing interactions.
  • Use GPT-5 for complex use cases such as technical support, multi-step workflows, or conversations that require deep reasoning.
  • Use GPT-4o if speed is critical and the conversations are relatively straightforward (e.g., FAQ-style responses).
  • Test before publishing. After changing the model, use the Preview tab to test how your agent responds. Different models may interpret the same instructions slightly differently.