Skip to content

Language Model

Language Model is a language model where the provider is just another field: you choose Provider (OpenAI, Anthropic, Google, Deepseek, Groq, Mistral, Ollama…) and that provider’s model. Inside, it works exactly like that provider’s own component, with the same answers, the same errors and the same spend.

The difference is that switching provider does not mean switching node. You can try the same flow with several models from the canvas, or let the application that calls the flow choose the provider on each request, without touching the deployed version.

It is used like the other models: by connecting its Model output to an agent’s Model input, or on its own, to answer the text it receives on Input.

When to use it

  • The application that uses your flow must be able to choose the model on each call, for example a cheaper one for basic-plan customers and a better one for advanced-plan customers.
  • You want to compare several providers with the same flow before settling on one.
  • You want a backup provider ready and to switch to it through the API if the usual one fails or runs out of quota.

If the provider will never change, use its own component, such as Anthropic or OpenAI: it shows all its fields directly.

Configuration

In the Common tab:

FieldWhat it does
ProviderThe model’s provider. Default: OpenAI. Choose it first: when you change it, the node asks for that provider’s credential and loads its models.
Model NameThe provider’s model. Click the refresh button next to the field to load the models your credential can use, or type its name.
TemperatureHow much variety the answers have: low values make them steadier; high values, more creative. Models that do not take it ignore it. Default: 0.1.
Credential (required if the provider uses one)The credential of the chosen provider. The list only shows that provider’s credentials. Local models, such as Ollama or LM Studio, need none. See Credentials.
ConnectionWhat to connect
InputThe text the model should answer when it works on its own, for example the message from a Chat Input or the result of a Prompt.

The node remembers the credential you choose for each provider. If you choose OpenAI with its credential, switch to Anthropic and choose its credential, when you go back to OpenAI its credential comes back by itself. That is what lets you switch provider through the API without sending any key.

Advanced options

In the Advanced tab:

FieldWhat it does
System MessageGeneral instructions for the model when it works on its own: its role, the tone, what it should avoid.
StreamShows the answer in the chat as it is written. Only has an effect in the chat. Default: off.
Max TokensThe maximum length of the answer, in tokens. With 0 the provider’s own is used. Only shown if the provider supports it. Default: 0.
ReasoningHow much the model thinks before answering: default, none, low, medium, high or max. default leaves the same as the provider’s own component. The others are translated to the closest level the provider accepts. Only shown if the provider supports it.
Base URLThe address where a model installed on your servers answers, such as Ollama or LM Studio. Only shown with those providers.
Provider SettingsOther fields of the provider’s own component, each with its own name. For example, prompt_caching = false with Anthropic. Fields the provider does not have are ignored.
Credentials by ProviderThe credential the node remembers for each provider. It fills itself in when you choose a credential; you usually do not need to touch it.

Each provider’s fields that do not appear here keep their default value. If you need to change one, write it in Provider Settings.

Batch processing

With the Batch Mode switch on, the component runs once for each item of the list you connect to Batch Input and returns all the results together in a list. Parallel Executions sets how many items are processed at once (from 1 to 5). More in Key concepts.

Switching the provider through the API

When you call the flow from another application, change the provider and the model with tweaks, using the node’s ID:

{
"input_value": "Summarise this contract",
"tweaks": {
"MultiProviderModel-k3x9a": { "provider": "Anthropic", "model_name": "claude-sonnet-5-5" }
}
}
  • Always send the model together with the provider: an OpenAI model does not exist at Anthropic.
  • The provider must have been set up on the canvas before, with its credential, so that the node remembers it. Otherwise, send credential_id too.
  • The change only applies to that call. The deployed version does not change.

More in Changing fields for one call.

Output

OutputWhat it contains
ModelThe model, already configured. Connect it to the Model input of an agent or of any component that asks for a model. It also works as the model of a Voice Agent.
TextThe model’s answer to the text received on Input.

As with the other models, the two ways of using it exclude each other: if you connect Model, Input and Text are hidden, and the other way round.

Example

An online academy offers a help assistant on two plans. Its application calls the same flow for every student:

  1. On the canvas, add Language Model and connect Model to the agent.
  2. Choose Deepseek in Provider, its credential and a model. Then choose Anthropic, its credential and a model. The node now remembers both credentials.
  3. Leave Deepseek as the default provider and deploy the flow.
  4. For advanced-plan students, the application adds "tweaks": {"<node id>": {"provider": "Anthropic", "model_name": "claude-sonnet-5-5"}} to the call.

Good practices

  • Set up on the canvas every provider the application will use, even if you then leave only one as the default.
  • The spend is recorded on the credential of the provider that answers, and its limits apply as for any other node.