Inference API pricing
Pay 25% of the original official reference price, a 75% reduction. Eligible models include verified open-source models, Kimi K3 under its restricted open-weight licence, and GPT-5.6 Luna as the sole proprietary exception. This is not a discount on Edelux catalogue prices.
Check the source before you send a token
A model qualifies only when it appears in the live Edelux catalogue and has verified official prices and a licence reference. Kimi K3 has open weights with licence restrictions; it is not an MIT-licensed open-source model. GPT-5.6 Luna is proprietary, not open source; its reference links to the publisher’s model documentation. Models without verified pricing remain unavailable.
Your API account shows input, cached-input, and output prices in USD per 1 million tokens, beside the official reference and its verification date. Each request freezes the official price tiers and time-dependent rates at reservation. We reserve the bounded maximum, then use provider-reported input tokens to select the applicable frozen tier at settlement.
GPT-5.6 Luna uses the publisher’s Standard rates. Above 272,000 input tokens, the higher input tier applies to the full request, up to 922,000 input tokens. Cached-input rates cover cache reads; this API does not offer separately metered cache writes.
Sign in to see the verified models and rates available to your account. Personal Chat and API access share a limited model selection; business accounts retain their broader selection. This public page does not publish a separate model list or estimated prices.
Personal customers can manage their API account directly. New Business accounts require contacting sales@edelux.studio.
A separate USD balance
Pay with crypto through NOWPayments to add $10, $50, or $100 USD of prepaid, nonwithdrawable API credit. Checkout availability appears in your account. NOWPayments quotes the crypto amount and network; network fees and confirmations apply.
We add only the purchased USD credit after verifying the completed payment. Underpaid invoices add no credit; overpayment adds no extra credit. Each top-up requires a payment you initiate, never an automatic wallet debit.
Dedicated fapi_ keys spend this API balance only. Personal workspace owners and business organization owners or admins can set an expiry or lifetime USD cap and revoke a key at any time.
API credit has no monthly reset and stays separate from monthly Build and Chat subscription credits and marketplace funds. Subscription credits do not pay for API requests.
Use your existing chat client
The API supports OpenAI-compatible text and tool chat, including streamed responses. Your account provides the full base URL and a curl example using an eligible model.
POST /inference/v1/chat/completionsBefore dispatch, the account reserves the maximum request cost. Settlement uses provider usage when available; if usage is missing, the authorized maximum applies and the account marks it as estimated.