1. How do I use AI Agent Studio?
  2. Large Language Models (LLMs) in AI Agent Studio

Large Language Models (LLMs) in AI Agent Studio

AI Agent Studio supports LLMs for workflows and nodes.

You can select the model considering the model type that best matches the capability, performance, and cost requirements of your workload. AI Agent Studio provides the following model types:
  • Basic: Supports standard workloads for which cost is more important than peak capability. Example workloads include summarization, question answering, retrieval-augmented generation (RAG), and lightweight reasoning.
  • Balanced: Supports workloads that require a balance of quality, latency, and cost. Example workloads include RAG, tool use, multistep reasoning, and agent orchestration.
  • Frontier: The highest-capability models, designed for complex and long-running workloads. Example workloads include autonomous execution, deep research, complex reasoning, and policy model creation.

Models Available in AI Agent Studio

You can review the models available in AI Agent Studio, including configured Bring Your Own LLM (BYOLLM) models, from the Credentials tab in AI Agent Studio. Open the LLM subtab to view the Oracle-provided models and any additional models that you have added. For more information, see Add Your LLM
Note: The models available for each model type can vary by release.

How to Define a Model in a Workflow

You can define a model at the workflow level and also at the node level for the LLM, Agent, and Document Processor nodes within the workflow. The model defined at the workflow level is used by all applicable nodes in the workflow unless a different model is specified in the node settings.

  • To select a model at the workflow level, go to Workflow Settings and select the LLM subtab. Select the model from the Model configuration list.
  • To select a model at the node level, go to the settings for an LLM, Agent, or Document Processor node and select the model from the Model list.

Deprecated Models

As newer models become available for selection in AI Agent Studio, some older models will be deprecated and eventually retired. When a deprecated model is configured for a workflow or node, a deprecation warning is displayed. The deprecation status is also shown on the LLM Card. You can continue to use a deprecated model, but it’ll become unavailable when it’s retired. If you’re using a deprecated model, we recommend that you select one of the available models instead.

Supported Models

The following table lists the supported models by model type and status. The model identified as default is preselected by default when you configure a model.

Supported Models

Model Type Model Status
Basic GPT-OSS 120B Supported
Balanced GPT-4.1 mini Deprecated
Balanced Gemini 3.1 Flash-Lite Supported
Balanced GPT-5 mini Deprecated
Balanced GPT-5.6 Luna Default
Balanced GPT-4o mini TTS Supported
Balanced GPT-4o Transcribe Supported
Balanced GPT-4o mini Transcribe Supported
Balanced GPT-4o Transcribe Diarize Supported