Language Model
On this page
Purpose
Language Model sends an input and optional system message to a chat model. Use it for one model request and one response when the flow does not need tool selection or agent planning.
When to use it
- Generate, classify, summarize, extract, or rewrite text.
- Add a model response before Chat Output or Text Output.
- Use Agent Core instead when the model must decide whether to call tools.
Required setup
Arkus is the default provider and uses the platform-managed gpt-4.1 deployment. It does not show an API key field. OpenAI, Anthropic, and Google each require their matching API key after you select that provider.
How to use it
- Add Language Model and connect a text or Message value to Input.
- Select a provider and model.
- Add a System Message when the response needs a defined role or format.
- Connect Model Response to Chat Output, Text Output, or another Message-compatible input.
Configuration
| Setting | What it controls | Recommended starting point |
|---|---|---|
| Model Provider | Selects Arkus, Anthropic, Google, or OpenAI. | Start with Arkus. |
| Model Name | Selects the provider model or Arkus deployment. | Keep the default Arkus deployment, or choose a model available to the selected provider account. |
| Input | The text sent to the model. | Connect a Message-producing component. |
| System Message | Adds instructions that guide the response. | Leave blank for simple prompts. |
| Stream | Streams the response when enabled. | Leave off until the downstream path needs streaming. |
| Temperature | Controls response variation. | Start at 0.1. |
Expected result
Model Response invokes the model and returns a Message. Language Model returns the configured model object without invoking it, for components such as Structured Output. External providers fail configuration when their required key is missing.
For OpenAI model names beginning with o1, Arkus hides System Message. The provider's model list can change, so treat the visible choices as the current options, not a complete catalog.
Reference details
The source-defined OpenAI reasoning models ignore Temperature. Use streaming only when the receiving component supports partial messages.
Provider credentials and credits
Arkus model calls use Arkus credit metering. Selecting OpenAI, Anthropic, or Google uses that provider account instead. For a deployed flow, store the external key as a saved Credential reference. Do not place a key in a prompt or message field.
Related components
- Chat Output displays a model response in the Playground.
- Text Output returns a model response through the Platform flow API.
- Structured Output uses the model object for schema-guided extraction.