Verification and deployment limitations
The public billing catalog tells you which model identifiers have configured prices. It does not independently establish successful inference, upstream identity, context limits or support for tools, vision and structured outputs on every route.
Which source should I use?
| Question | Source | Limitation |
|---|---|---|
| What is the current configured USD rate? | Live catalog | A billing listing is not a route health check. |
| Which endpoint accepts a request? | API overview and OpenAPI | Contract syntax validation does not prove production behavior. |
| What can a model do? | Relevant upstream documentation and a permitted integration test | Provider capability does not automatically imply gateway support. |
| What is retained? | Data handling | Production infrastructure settings need operator verification. |
Pricing snapshots
Static prices were checked against the public billing endpoint on October 21, 2026. They are expressed in USD and retain the exact billing unit. AIx chat billing in the inspected implementation applies a single rate to billed tokens. Do not read that rate as separate input and output prices from the upstream provider.
Example test levels
A syntax check verifies that a program parses. A mocked contract test verifies the request and response handling against a controlled fixture. Neither is a live model benchmark. Authenticated inference, latency and production billing tests require approved credentials and a permitted budget; they have not been performed here.
The repository’s examples/ programs use environment variables for secrets. The
Python walkthrough explains setup and failure handling.
Review and corrections
Recheck pricing before purchase, review API guides after backend changes, and report contradictions through support. Documentation does not provide an independent guarantee of provider retention or model availability.