Effective 2026-08-21 · applies to the inference API at api.jetinfer.com
Sold through OpenRouter. You pay only for tokens we successfully serve, we never train on your data, and we do not promise an uptime SLA yet. Either side can walk away at any time.
JetInfer is operated by Quantive Bt., a company registered in Hungary. “We” and “us” mean Quantive Bt.; “you” means the person or company using the API. Reach us at hello@jetinfer.com.
An OpenAI-compatible inference API serving Qwen3.8-27B, quantized to int4
(W4A16). The quantization is disclosed on the landing page and in the
/v1/models response, because it affects output quality and you
should know what you are buying.
We may change models, rates, or capacity. Because OpenRouter holds the customer relationship, changes reach you through them: a rate change is published to our provider listing before it takes effect, and withdrawing a model removes the endpoint from it. You are never charged a rate that was not published before your request.
Per token, at the rates on our OpenRouter provider listing and in
/v1/models. Those two are generated from a single source and are the
authoritative prices. This site does not repeat them: rates move with the market,
and a copy kept here would eventually contradict the rate you are actually
charged. What bills how is set out on the
specifications page and does not change with the rate.
You are never billed for a request we failed to serve. A request that returns an error, or for which we cannot account for token usage, is not charged. Settlement runs through OpenRouter under their provider terms: they hold the billing relationship with you, we invoice them against reconciled token counts, and we never see or store your payment details.
These are the limits of what we owe you:
429, not a place in a queue.The service is provided “as is”, without warranties of any kind so far as the law allows. Our total liability for any claim is limited to the amount you paid us in the three months before the claim arose.
Do not use the API to:
We may refuse or suspend traffic doing any of the above, and where it is safe to do so we will say why through OpenRouter first. Because OpenRouter holds the billing relationship, you are charged for served tokens only: there is no balance with us to strand, and anything already paid for tokens we did not serve comes back through their reconciliation.
We never store prompts or outputs, and we never train on your data. The detail is in our privacy and data retention policy, which forms part of these terms.
You keep every right in what you send and in what the model returns to you. We claim no licence over either.
You can stop sending us traffic whenever you like — deselect us in OpenRouter and nothing else is owed. We can stop serving with notice, or immediately where the law requires it or in response to abuse. Either way you are charged for served tokens only, so there is nothing to unwind.
When these terms change we update this page and the effective date above. Material changes are published here before they take effect, and where a change affects how requests are served or billed it is reflected on our OpenRouter provider listing at the same time. These terms are governed by Hungarian law and the courts of Hungary have jurisdiction, without affecting consumer protections you hold where you live.
Questions about these terms, about billing, or about anything else: hello@jetinfer.com.