Choosing the right model

In your Agent configuration, you can select from models made available by your organization through CARTO AI settings.

Model recommendations

Models generally fall into two categories:

• Opus/Pro models (e.g., claude-opus-4.8, gemini-3.1-pro, gpt-5.5-pro): Most capable, best for complex multi-step reasoning. Highest cost and latency.

• Sonnet/Pro models (e.g., claude-sonnet-4.6, gemini-3.1-pro, gpt-5): Strong balance of capability, cost, and latency. claude-sonnet-4.6 is the default CARTO-managed model.

• Flash/Mini/Haiku models (e.g., gemini-3.5-flash, gpt-5-mini, claude-haiku-4.5): Faster and more cost-efficient. Good for simpler tasks.

If the same model is available through multiple cloud providers, choose based on your data residency requirements.