FAQ

Integration

I already use the OpenAI SDK — how much has to change?

Two settings: point base_url at https://api.apione.ai/v1 and use an APIONE key. Use the model IDs listed on APIONE (e.g. gpt-5.6-sol). Nothing else changes.

Can I use both the OpenAI and Anthropic protocols?

Yes — one key serves both. Which you pick depends on your SDK and tooling; see OpenAI compatibility and Anthropic compatibility.

Are function calling, structured output and vision supported?

Yes, where the model supports them. Each model's capability matrix is on its page in the model plaza; sending an unsupported parameter returns 400.

Can I call the API straight from the browser?

Technically yes — but don't. The key would be visible to every visitor. Have the browser call your backend, and let the backend hold the key.

Billing

Is there a monthly fee or minimum spend?

No. Pay as you go means topping up and paying for actual usage, and balance does not expire; subscriptions are an optional discount — see billing.

Do failed requests cost anything?

4xx responses and 5xx responses that produced no content are free, as are failed failover attempts. A stream interrupted midway does bill the tokens already produced — see billing.

Is it more expensive than buying direct?

Every model's actual rate is listed in the model catalog, so you can compare it with the vendor's own price.

Reliability

Does a vendor outage take me down?

Most models have several providers, and the gateway moves the request within the same call — see failover. Only a simultaneous failure across every provider for that model reaches you as an error.

Why do two identical requests differ so much in latency?

They may have hit different providers, or upstream was queueing. Search the request ID (the X-Maxx-Request-ID response header) in console › Usage › Request log to see each call's first-token latency; if it stays slow, contact us with a few request IDs.

Is there an SLA?

Enterprise customers who need a written SLA should contact us to discuss terms.

Compliance and data

Do you store my requests?

Request and response bodies are not kept after the request by default; only token counts and metadata needed for billing are. See the privacy policy.

Which third parties see my data?

Only the upstream model provider that served the call, plus the cloud, network, payment and support vendors needed to run the service. See the privacy policy.

Can I choose the processing region?

Requests can't pin a provider or region; call content goes to the region of the upstream provider for the model you choose. If you have hard compliance requirements, contact us to confirm.