On this page

Track the two limit families

xAI documents prepaid credits and optional monthly invoiced billing. It also applies per-model request and token rate limits. Your usage dashboard should therefore separate account spend from request throughput.

For spend, watch credit balance, daily cost, model, and API key. For throughput, watch requests per second and tokens per minute, plus any HTTP 429 responses.

Include every token type

xAI notes that prompt, completion, reasoning, and cached prompt tokens can count toward throughput. Cached tokens may be billed at a lower rate while still contributing to a token-per-minute ceiling.

That distinction explains why cost can look efficient while a high-volume workflow still reaches a rate limit.

Keep your usage visible while you work. Get Super Notchy for $27 ↗

Use the console and the local session

The xAI Usage Explorer can group consumption by API key, model, token type, and time. Use it for the account-wide view. Local application logs explain which job generated the traffic.

Super Notchy can read the usage exposed to the signed-in Grok CLI and show the headline credit status on the Mac. Keep the console as the detailed billing authority and the local reading as the interruption warning.

Sources and further reading