Model catalog

OpenAI-compat GET /v1/models with a modality field that names text / image / video.

Endpoint

GET https://api.genie.tech/v1/models
Authorization: Bearer sk-genie-{your-key}
Content-Type: application/json

Try it

Try it nowList available models

Response

{
  "object": "list",
  "data": [
    { "id": "qwen3:0.6b",      "object": "model", "modality": "text",  "owned_by": "genie" },
    { "id": "qwen3:14b",       "object": "model", "modality": "text",  "owned_by": "genie" },
    { "id": "qwen3-coder:30b", "object": "model", "modality": "text",  "owned_by": "genie" },
    { "id": "flux-schnell",    "object": "model", "modality": "image", "owned_by": "genie" },
    { "id": "sdxl-turbo",      "object": "model", "modality": "image", "owned_by": "genie" },
    { "id": "wan-2.1",         "object": "model", "modality": "video", "owned_by": "genie" },
    { "id": "ltx-video",       "object": "model", "modality": "video", "owned_by": "genie" }
  ]
}

Sources of truth

Text models = union of loadedModels across online runners that advertise jobType='inference'. Empty if no inference runner is up — listing models nothing serves would mislead.
Image models = curated whitelist matching the POST /v1/images/generations Zod enum.
Video models = same pattern.

The catalog is non-sensitive but auth-gated so per-org filtering (free-tier model clamps, allowedModels per key) can apply.