Models & pricing

Every model is sold at its list price: the model maker's own published price. Everyone pays it; holders earn cash back instead of a lower price.

The catalog has 248 models. Browse and filter them on the models page, or list them with GET /v1/models.
Ids are provider/model, e.g. openai/gpt-5.4 or meta/llama-3.1-8b-instruct-fp8.
Prices are per unit: per 1M input and output tokens for text, per image, per second of video, per 1K characters or per minute of audio.
List price for everyone. There is no markup and no fee on top.
Cash back, not discounts. Holders pay the same list price. Pro and Elite accounts (Pro: 1,337 $INFX in a linked wallet; Elite: 1,337 $XAVA + 1,337 $INFX) get a bigger buyback on every spend, and anyone can take up to 5% back in USDC, $INFX or another supported token.
Prices refresh hourly. The price that applies is the one at the time of the request.

GET /v1/models (one entry)

{
  "id": "openai/gpt-5.4",
  "object": "model",
  "owned_by": "OpenAI",
  "zdr_supported": true,
  "x_elite_discount_bps": 0,
  "x_pricing": {
    "input_tokens":  { "unit": "per 1M tokens", "list_usd": "2.50", "elite_usd": "2.50" },
    "output_tokens": { "unit": "per 1M tokens", "list_usd": "15.00", "elite_usd": "15.00" }
  }
}

In x_pricing, list_usd is the list price everyone pays. elite_usd and x_elite_discount_bps are kept for compatibility: there are no holder prices, so they always equal the list price and 0. The shape is a draft.

zdr_supported is true when the model keeps zero data retention end to end: neither we nor the model provider store your prompts or outputs. On the models page these carry a ZDR supported badge and a filter.