Status

What is checked. The chat checks send a one-line prompt with thinking switched off to a model on each upstream, plus GPT-OSS 120B separately because it is the most used; they pass when text comes back. The Responses API check passes on a completed response. Embeddings must return a 1,024-dimension vector, rerank must put the right passage first, text to speech must return audio, and speech to text must transcribe that audio correctly. Probes run on the API server itself, so a network problem between you and us would not show here. A single failed probe is usually a busy backend and clears on retry; two or more in a row are listed as an incident below.

If your requests fail while everything here is green, check the error docs first: 401 and 402 are account-side, and 404 model_not_found means the id is not in the catalog.