Skip to content
Docs

Models

List the models the API serves, with their capabilities and prices in credits.

Updated

GET /models returns every model you can call through the API, in the OpenAI list format with a few extra fields. The same catalog, with filters, is on the Models page.

curl https://ewpire.com/api/v1/models \
  -H "Authorization: Bearer $EWPIRE_API_KEY"

A model in the list looks like this. The entry below is taken from the live list when this page is generated, so its prices are the ones the API returns today:

JSON

{
  "id": "anthropic/claude-sonnet-5",
  "object": "model",
  "created": 0,
  "owned_by": "Anthropic",
  "name": "Claude Sonnet 5",
  "description": "…",
  "context_length": 1000000,
  "capabilities": [
    "text",
    "reasoning",
    "vision",
    "files",
    "tools",
    "coding"
  ],
  "input_modalities": [
    "text",
    "image",
    "file"
  ],
  "output_modalities": [
    "text"
  ],
  "pricing": {
    "input_credits_per_million_tokens": 1.94,
    "output_credits_per_million_tokens": 9.7
  }
}

A model that charges long requests at a higher rate carries that rate in pricing.long_context. Again from the live list:

JSON

{
  "pricing": {
    "input_credits_per_million_tokens": 2.482353,
    "output_credits_per_million_tokens": 12.411765,
    "long_context": [
      {
        "min_prompt_tokens": 272000,
        "input_credits_per_million_tokens": 4.964706,
        "output_credits_per_million_tokens": 18.617648
      }
    ]
  }
}

Fields

FieldMeaning
idWhat you pass as model.
owned_byThe lab that makes the model.
context_lengthThe most tokens the model takes in one request.
capabilitiesWhat it is good at: reasoning, coding, vision, tools, files.
input_modalitiesWhat it reads: text, image, file.
pricingCredits per million input and output tokens.
pricing.long_contextOnly on a model that charges long requests at a higher rate, lowest threshold first. Once a request's input reaches min_prompt_tokens, the whole request is charged at that entry's rates, not only the tokens past it. See Usage.

created is always 0: the catalog does not keep release dates.

Auto

The first entry is auto. With "model": "auto", ewpire picks a model for each request and, if that model is unavailable, tries the next one. The model field of the answer names the one that answered.

Picking a model

  • Only models that answer in text are served through the API. Image, video and music models work in the app; see Images and Video.
  • To send images in a message, pick a model with image in input_modalities.
  • Models come and go as the catalog changes, without notice; auto is never removed. Read GET /models rather than hard-coding a long list, and handle 404 model_not_found.