# Debug models

The Debug page lets you choose a model, endpoint, and input, then inspect the result immediately. Before sending a request, prepare an API Key on **Keys** that can access the selected model.

## Steps

1. Open **Debug** in the left sidebar.
2. Choose an endpoint tab for the capability you want to verify, such as chat (text), chat (multimodal), Messages, Responses, embeddings, rerank, OCR, text-to-image, text-to-video, TTS, or ASR.
3. Select a model and fill in the inputs required by the endpoint. Text, images, queries, and candidate documents use different controls; follow the page.
4. Adjust available parameters as needed. Common chat parameters include temperature, frequency penalty, presence penalty, max output tokens, seed, streaming, and system prompt.
5. Click **Send**, **Run**, or the submit action for the current endpoint.
6. Review the returned result and request information. Endpoints may show output, request body, raw response, Token usage, latency, or similarity scores.
7. To start over, use **Clear conversation** or **Clear results** when the endpoint provides them.