Does my OpenAI code work?
Yes for chat completions, the stateless Responses API and legacy completions: change the base URL and key. See Requests and outputs for what is passed through.
How is Zalmir different from calling a provider directly?
One key and one balance for many models, routing and fallbacks across endpoints, and a signature check on every answer from an independent seller before you receive it.
What does verified mean?
The answer was signed by the seller's key and the signature checked. It proves who sent the bytes, not which model produced them.
Who can see my prompts?
For independent-seller endpoints, the seller can see prompts. Details in Privacy controls.
Am I charged for failed requests?
No. You are charged only for an answer you receive after its checks pass. Canceled and failed requests release their reservation.
Can I get a refund for credits?
No. Credits are prepaid and non-refundable.
What does Zalmir charge?
No markup on provider models. A fee on each top-up: 2.5%, falling to 1% as your total top-ups grow. See Pricing.
Are there free models?
Not yet. :free returns 404 no_free_endpoints.
Does streaming send tokens as they are generated?
No. Zalmir checks the whole answer first and then streams it.
Are embeddings supported?
Not yet.
Can I use my own provider account?
Yes, with $0 Zalmir fee. See Your own provider keys.