Docs / Models

Models and pricing

Three models, one price list. All three are abliterated: the refusal behaviour has been removed from the weights, so they answer requests a stock model would decline.

Catalog

Models
Model idNameUse it for
model-zeroModel ZeroFast, low-cost reasoning tuned for agent loops, tool calls and high-volume pipelines.
model-one-flashModel One FlashEfficient general-purpose inference for everyday production traffic.
model-oneModel OneOur larger model for advanced reasoning and long, complex tasks.

Send the model id in the model field. The same id comes back in every response. Context windows and speeds are not published yet.

Prices

US dollars per 1 million tokens.

Prices per 1M tokens
Model idInputCached inputOutputBonus credits
model-zero$0.50$0.10$1.00Accepted
model-one-flash$0.40$0.10$1.00Accepted
model-one$1.50$0.30$3.00Paid credits only

Worked example

A model-zero request with 12,000 prompt tokens, 10,000 of them cached, and 800 output tokens:

Cost
input   2,000 × $0.50 / 1M = $0.001000
cached 10,000 × $0.10 / 1M = $0.001000
output    800 × $1.00 / 1M = $0.000800
total                        $0.002800

List models

GET /v1/models returns the models that are currently enabled, sorted by id. It needs a valid key but does not touch your balance.

cURL
curl https://api.hmbrd.dev/v1/models \
  -H "Authorization: Bearer $HUMMINGRELAY_API_KEY"
Response
{
  "object": "list",
  "data": [
    { "id": "model-one", "object": "model", "created": 1762560000, "owned_by": "hummingrelay" },
    { "id": "model-one-flash", "object": "model", "created": 1762560000, "owned_by": "hummingrelay" },
    { "id": "model-zero", "object": "model", "created": 1762560000, "owned_by": "hummingrelay" }
  ]
}

There is no GET /v1/models/{id}. A model id that is unknown or disabled answers 404 model_not_found on completion requests.