Frequently asked questions
Common integration, billing and troubleshooting questions.
Which models can one key use?
A key can call models allowed by its account, group and key permissions. Copy exact IDs from pricing. The client must support the protocol and model capabilities.
Why am I getting 404?
Check for a console URL, duplicated /v1, or a complete endpoint entered as the SDK base. See Endpoints and Base URL.
Why does a request fail with credit remaining?
Check whether the key is disabled, expired, exhausted or missing model permission. Account credit and key allowance are separate limits. Inspect the error code.
Is caching always cheaper?
Caching requires model support and actual cache usage. Reads and writes may have separate rates. Settlement uses published model rates and actual usage; see Billing.
Does credit expire? Are refunds or invoices available?
Expiry, refunds, invoices and payment methods follow formal information shown before payment. Ask support if these details are missing. Do not assume permanent credit or unconditional refunds.
Can I retry a content refusal?
Revise the content and read the Content policy. Repeated refusals can trigger restrictions. Do not rotate keys to bypass them.
What should I send support?
Provide the local request ID, time, model ID and redacted error code. Never send a complete key or sensitive prompt.
Still have a question?
Support details have not been published.