The versioned API base the chat client posts /chat/completions under: the configured root plus /v1, as the model lister's /v1/models is.
The versioned API base the chat client posts /chat/completions under: the configured root plus /v1, as the model lister's /v1/models is.
A base URL that already ends in /v1 is used as it is, so the https://api.mistral.ai/v1 that Mistral's own OpenAI-SDK examples use works too. The old client appended /v1 unconditionally and so posted to /v1/v1/chat/completions for such a URL.