For OutSystems AI Agent development in ODC, there isn’t a single “best” model; the platform is explicitly model‑agnostic. The “most suited” model depends on your use case and constraints.

OutSystems supports, out of the box, at least these categories of models for AI Agents:

1. Azure OpenAI models (GPT family)

   • Natively integrated in AI Agent Builder.
   • Strong choice for: general chat agents, summarization, Q&A over documents, code‑oriented agents, and enterprise scenarios already on Azure. AI Agent Builder by OutSystems > Overview and Purpose
   • Often the default recommendation if your organization is already using Azure or has strict data‑residency / compliance needs tied to Microsoft. New GenAI Use Cases with OutSystems > Roshat AI Application > Architecture and Platform Choice

2. AWS Bedrock models (Anthropic Claude, etc.)

   • Also natively supported by AI Agent Builder. AI Agent Builder by OutSystems > Overview and Purpose
   • Strong choice if your enterprise is AWS‑centric, or you prefer specific Bedrock models (e.g., Claude) for reasoning‑heavy agents, long‑context document work, or stricter data policies via AWS. Agent Workbench > AI Model Connections in Agent Workbench

3. Other providers (OpenAI, Gemini, custom / on‑prem models)

   • Any model that exposes a REST API can be plugged in. AI Agent Builder by OutSystems > Overview and Purpose Agent Workbench > AI Model Connections in Agent Workbench
   • Good for: regulated environments that require self‑hosted models (e.g., Llama‑based), or when you want a niche capability not yet exposed via Azure/AWS.

So, “most suited” in practice is:

• If you’re on Azure → use Azure OpenAI GPT models as your primary choice. They’re first‑class citizens in AI Agent Builder and proven in reference apps like Roshat. New GenAI Use Cases with OutSystems > Roshat AI Application > Architecture and Platform Choice
• If you’re on AWS → use AWS Bedrock (e.g., Anthropic Claude) via AI Agent Builder. Building Better, Faster: AI Enhancements for OutSystems Apps > AI Agent Builder Overview and Workflow
• If you need custom/sovereign AI → connect your preferred LLM through REST as a custom model connection. Agent Workbench > AI Model Connections in Agent Workbench

When choosing between them, consider:

• Data & Compliance: keep data inside your existing cloud (Azure vs AWS vs self‑hosted).
• Use Case:
  • Chatbot / copilot / RAG over PDFs → GPT‑4‑class or Claude‑class models. New GenAI Use Cases with OutSystems > Roshat AI Application > PDF Interaction and Summarization
  • Heavy reasoning / long documents → larger, long‑context models (e.g., Claude via Bedrock or latest GPT variants).
• Cost & Latency: you can monitor token usage and adjust the model accordingly in AI Agent Builder. New GenAI use cases with OutSystems > AI Agent Builder by OutSystems > Technical Features and Integration

Summary:
Use Azure OpenAI GPT or AWS Bedrock (Claude, etc.) via AI Agent Builder, depending on your enterprise cloud and requirements; both are “most suited” and officially supported for OutSystems AI agents, with other LLMs available via REST when needed.