Model and parameters

Important: Agents is currently in private beta and is not yet available to all customers.

A model defines the AI provider and model used by the agent to generate responses. The selected model determines how the agent processes requests and generates output. The available parameters depend on the model provider and the selected model.
The following models are available. Each model may include distinct parameters based on the provider and the selected model.

  • AWS ChatBedrock

  • Azure ChatOpenAI

  • ChatAnthropic

  • ChatCohere

  • ChatGoogleGenerativeAI

  • ChatMistralAI

  • ChatOpenAI

  • ChatOpenRouter

Model parameters define how the selected model generates responses. The available parameters depend on the model provider and the selected model. For more information, seeModel parameters.

Model parameters

Field or selection Description Variable types
Connection The connection used to authenticate with AWS Bedrock. N/A
Region The AWS region where the Bedrock service and model are hosted. N/A
Model name The model identifier used by AWS Bedrock. The available model names depend on the selected region and provider. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses. Decimal, Integer
Max tokens to sample The maximum number of tokens the model can generate in a response. This setting may not apply to all models. Decimal, Integer
Add field

Add optional model parameters supported by the selected model. Available fields depend on the model.

Custom model name: Specify a custom model name.

 Text, Collection
Agent role and rules Define the agent’s role, behavior, and constraints. Text
Work request Describe the task or question the agent should respond to. Text
Field or selection Description Variable types
Connection The connection used to authenticate with Azure OpenAI. N/A
Model name The model identifier used by Azure OpenAI. The available model names depend on the selected region and provider. N/A
Reasoning effort The level of reasoning effort used by the model. Select Low for faster, simpler responses, Medium for a balance of quality and speed, or High for more thorough responses. This setting applies only to supported reasoning models. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses. Decimal, Integer
Add field

Add optional model parameters supported by the selected model. Available options include:

  • Max tokens: The maximum number of tokens the model can generate in a response.

  • Top probability: Token selection by sampling from the probability distribution of likely tokens.

  • Frequency penalty: Reduces repetition by penalizing tokens that appear frequently in the generated text.

  • Presence penalty: Encourages the model to introduce new topics in the response.

  • Timeout: The maximum time allowed for the model request to complete.

  • BasePath: The base path used for the API request.

  • BaseOptions: Additional base configuration options used when sending the request to the model.

 Text, Decimal, Integer, Boolean, DateTime, Collection
Field or selection Description Variable types
Connection The connection used to authenticate with Anthropic. N/A
Model name The model identifier used by Anthropic. The available model names depend on the selected region and provider. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses. Decimal, Integer
Max tokens to sample The maximum number of tokens the model can generate in a response. This setting may not apply to all models. Decimal, Integer
Budget tokens The maximum number of tokens that can be used for reasoning before generating the final response. Decimal, Integer
Add field

Add optional model parameters supported by the selected model.
Available options include:

  • Max tokens: The maximum number of tokens the model can generate in a response.

  • Top P: Token selection by sampling from the probability distribution of likely tokens.

  • Top K: Limits token selection to the top K most likely tokens.

  • Extended thinking: Enables additional reasoning steps before generating the response.

 Text, Decimal, Integer, Boolean, DateTime, Collection
Field or selection Description Variable types
Connection The connection used to authenticate with Cohere. N/A
Model name The model identifier used by Cohere. The available model names depend on the selected region and provider. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses. Decimal, Integer
Field or selection Description Variable types
Connection The connection used to authenticate with Google Generative AI. N/A
Model name The model identifier used by Google Generative AI. The available model names depend on the selected region and provider. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses.. Decimal, Integer
Add field

Add optional model parameters supported by the selected model.
Available options include:

  • Custom model name: Specify a custom model name.

  • Max output tokens: The maximum number of tokens the model can generate in a response.

  • Top probability: Token selection by sampling from the probability distribution of likely tokens.

  • Top next highest probability tokens: Limits token selection to the top set of most likely tokens.

  • Safety settings: Configure safety filters applied to the model response.

  • Base URL: The base URL used for the API request.

 Text, Decimal, Integer, Boolean, DateTime, Collection
Field or selection Description Variable types
Connection The connection used to authenticate with Mistral AI. N/A
Model name The model identifier used by Mistral AI. The available model names depend on the selected region and provider. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses. Decimal, Integer
Add field

Add optional model parameters supported by the selected model. Available options include:

  • Max output tokens: Defines the maximum number of tokens the model can generate in a response.

  • Top probability: Token selection based on the probability distribution of likely tokens.

  • Random seed: The seed value used to generate deterministic responses.

  • Safe mode: Enables additional safety filtering for the generated response.

  • Override endpoint: A custom endpoint used for the API request.

 Text, Decimal, Integer, Boolean, DateTime, Collection
Field or selection Description Variable types
Connection The connection used to authenticate with OpenAI ChatGPT. N/A
Model name The model identifier used by OpenAI. The available model names depend on the selected region and provider. N/A
Reasoning effort The level of reasoning effort used by the model. Select Low for faster, simpler responses, Medium for a balance of quality and speed, or High for more thorough responses. This setting applies only to supported reasoning models. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses. Decimal, Integer
Add field

Add optional model parameters supported by the selected model. Available options include:

  • Max tokens: Maximum number of tokens the model can generate in a response.

  • Top probability: Token selection based on the probability distribution of likely tokens.

  • Frequency penalty: Reduces repetition by penalizing tokens that appear frequently in the generated text.

  • Presence penalty: Encourages the model to introduce new topics in the response.

  • Timeout: Maximum time allowed for the model request to complete.

  • Strict tool calling : Enforces strict validation when tools are used by the model.

  • Stop sequence: Sequence that stops the model from generating additional tokens.

  • BasePath: Base path used for the API request.

  • Proxy URL: Proxy endpoint used when sending the request.

  • BaseOptions : Additional base configuration options used for the request.

 Text, Decimal, Integer, Boolean, DateTime, Collection
Field or selection Description Variable types
Connection The connection used to authenticate with OpenRouter. N/A
Model name The model identifier used by OpenRouter. The available model names depend on the selected region and provider. N/A
Temperature The creativity of the generated response. Lower values produce more predictable responses, while higher values produce more varied and creative responses. Decimal, Integer
BasePath Base path used for the OpenRouter API request. Text
Add field

Add optional model parameters supported by the selected model. Available options include:

  • Max tokens: Maximum number of tokens the model can generate in a response.

  • Top probability: Token selection based on the probability distribution of likely tokens.

  • Frequency penalty: Reduces repetition by penalizing tokens that appear frequently in the generated text.

  • Presence penalty: Encourages the model to introduce new topics in the response.

  • Timeout: Maximum time allowed for the model request to complete.

  • BaseOptions: Additional base configuration options used for the request.

 Text, Decimal, Integer, Boolean, DateTime, Collection