At a glance
- 1,000,000 trial tokens per Organization, shared across your Tenants.
- Every request metered: tokens, timing and result.
- All of it visible in Usage and Activity. Usage is not a bill.
Your first million tokens
Every new Organization gets a trial credit of 1,000,000 tokens.
- Shared across your Organization. All your Tenants draw from the same credit.
- Input and output both count. Every token in and out of a model counts, including chat, completions, embeddings and reranking, on any model you deploy.
- What it doesn't cover. Trial tokens are spent on inference. Fine-tuning jobs don't use them.
- See it go down. Your Usage page shows the credit granted, used and remaining, and when it's likely to run out at your recent rate.
- When it runs out. Requests already running always finish. Rackspace can add further credit to your Organization.
How usage is metered
Every request to a model goes through the rackAI gateway, the only route to a model server. The gateway records one usage entry per request with:
- the Tenant, deployment, model or adapter, and API key that made it;
- input, output and cached tokens;
- time to first token, total time and the result.
Streaming responses are metered too, including a stream the client disconnects from partway. Usage entries contain counts and identifiers, not the text of your prompts or responses.
Where you see it
- Usage shows your Tenant's recent requests with tokens, timing and status, and lets you filter to failed requests. Each request links to its full trace.
- Activity shows every change in your Tenant, and every refused request, with who made it and when.
- The API gives the same data: usage records per Tenant, usage per Organization and the trial credit balance.
Usage is not a bill. It shows what you used, not what you owe.
Billing and pricing
Prices aren't published on this site.