Inference quickstart

Choose a model, make a direct inference request, and read its response

Make your first text-generation request with an Inworld API key and a model ID. You can call a model directly without creating a router.

Choose a model

Browse inworld.ai/models and select Text. Choose a model and copy its API model ID. The example below uses openai/gpt-5.4.

For other outputs, start with Image generation, Video generation or Decisions.

Set your API key

Create a Standard API key in Inworld Portal and copy the complete Base64 credential. Set it as-is; do not encode it again:

bash

export INWORLD_API_KEY="<your Base64 API key>"

Run these examples on your server. Browser and mobile clients should use a backend-minted token or a server proxy. See authentication recovery if a request fails.

Send a request

bash

curl --request POST \
  --url https://api.inworld.ai/v1/chat/completions \
  --header "Authorization: Basic $INWORLD_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "openai/gpt-5.4",
    "messages": [
      {"role": "user", "content": "Write a 50-word bio for a software engineer."}
    ]
  }'

To use another text model, replace model with its API model ID. The request follows OpenAI's Chat Completions format; see OpenAI compatibility for SDK setup.

Read the response

Read the generated text from choices[0].message.content. The response also includes the model that answered and token usage. Here is an abbreviated example; the generated text and token counts will vary:

json

{
  "model": "openai/gpt-5.4",
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "A software engineer who builds reliable applications..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 18,
    "completion_tokens": 50,
    "total_tokens": 68
  }
}

Next steps

Chat completions: streaming, tool calling, structured outputs and other text-model capabilities.
Image generation: generate an image, save it, and guide the next result with a reference.
Video generation: create a job, poll its status and download the video.
Decisions: get typed answers that your application can act on.
Request-level routing: add model or provider preferences and fallbacks to a request.

Create a reusable router

For traffic splitting, conditional routes or saved routing configurations, follow the Routing quickstart. Creating a router through the API requires Router Write permission; calling a model directly does not.