Call any model through one OpenAI-compatible API, at the price on its page. Switch model in the middle of a conversation and it carries on where the last one left off. Hand a question to alphaneural/auto and it starts on the cheapest model that can answer, moving up only when it has to, under a spend ceiling you set. Every run keeps a ledger of what it cost.
Starts on the cheapest model that can answer and moves up only when it has to.
The least expensive model here that still reads images and calls tools.
The longest context on the shelf, for whole repositories.
A middle-of-the-range model, for when the cheap one will not do.
Every one of these answers on the same OpenAI-compatible endpoint, at the price shown. Nothing to install and no per-provider account. The cheapest are listed first.
Routes each request to the cheapest model that can handle it — one that sees images if you send one, calls tools if you send tools, and fits your conversation — and moves up to a stronger model if that one fails. No fixed price: you pay for whichever models ran, at their listed prices, capped by max_spend (default $0.10 per request). Without a verify schema it escalates on errors and refusals only, not on answer quality.
3.2M open models and datasets, mirrored and readable here. Deploy any supported one to a GPU by the hour, or run the weights we already host through the same key as everything else.