FAQ
The questions people ask before their first request
Billing, compatibility, failure modes and what happens to your prompts — answered plainly, with a link to the policy that governs each one.
Open LLMs is an OpenAI-compatible API for open-weight models. Change the base URL and the model id, top up at least US$1.00, and pay per request. There is no subscription, and prompts are not used to train any model.
Questions
What is Open LLMs?
An OpenAI-compatible API for open-weight models. You point an OpenAI client at our base URL, use an Open LLMs key, and pay per request from a prepaid balance in US dollars. The request and response shapes are the ones you already use.
How do I switch from OpenAI?
Two lines. Change the base URL to https://api.openllms.co/v1, and change the model id to one from our catalogue. Your client library, request shape, response shape and streaming all stay the same.
Which models can I use?
Whatever is in the live catalogue at the moment you ask — it changes as models are added and retired. The models page lists every one with its price, context window and licence, and GET /v1/models returns the same list to your code.
Is there a free tier?
No. The minimum wallet top-up is US$1.00, and you pay only for the requests you run. There is no subscription, no monthly minimum and no plan to cancel. Promotional credit is granted from time to time; when you hold some it is spent before your paid credit and it carries an expiry date.
How is each model priced?
In the unit that matches what it does: chat and vision models per million tokens, speech-to-text per minute of audio, video per second of output. The pricing page shows every current rate, read from the same catalogue that serves the requests.
What happens when my balance reaches zero?
The API answers HTTP 402 with an insufficient_balance error until you add credit. Nothing is queued, nothing is charged, and your API keys stay valid.
Why did my first request return 503?
The first call to a model that has been idle can arrive while it is still starting. The response carries a Retry-After header: wait that long and retry, and the second call succeeds. A 503 with the code backend_saturated means the opposite problem — capacity is full rather than cold.
Do you train on my prompts?
No. We do not train, fine-tune or evaluate any model on the prompts you send to the API or on the responses returned to you, and we do not use them for advertising. Storing prompts at all is opt-in, and the privacy notice is explicit that when you do switch it on, an operator can read a stored snapshot to investigate support requests and abuse.
Do you charge GST?
No. The operator is not registered under GST, collects no GST, holds no GSTIN and issues no GST invoice. No part of any charge is described as tax.
Can I get a refund?
Unused paid credit can be refunded to the original payment method under the Refund and Cancellation Policy. There is no withdrawal to a bank account and no transfer between accounts. Credit you paid for does not expire.
Where each answer comes from
Note:
This page restates; it does not govern
Every answer above is a plainer wording of something already published here. When the two differ, the policy wins: billing policy, refunds, privacy notice, pricing, models and the API reference.