ModKie Hosted Models

ModKie-hosted models are ModKie-tuned workers (RunPod Serverless / Pod) exposed through the same public APIs as the rest of the Market.

Base URL (Jobs): https://modkie.com/api/v1
Auth: Authorization: Bearer YOUR_API_KEY

Jobs models use createTask + recordInfo. Credits are charged only on terminal success. Optional per-task callBackUrl or account webhooks.

Legacy aliases still accepted: open-models/wan-2.2-i2v, open-models/qwen-image-edit.

WAN 2.2 I2V

Image-to-video with ModKie production defaults.

curl -X POST "https://modkie.com/api/v1/jobs/createTask" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "modkie/wan-2.2-i2v",
    "input": {
      "prompt": "Gentle camera push-in, cinematic lighting",
      "image_url": "https://example.com/frame.jpg",
      "duration": 5,
      "width": 832,
      "height": 480
    }
  }'

Response:

{ "code": 200, "msg": "success", "data": { "taskId": "task_xxx" } }

Poll:

curl "https://modkie.com/api/v1/jobs/recordInfo?taskId=TASK_ID" \
  -H "Authorization: Bearer YOUR_API_KEY"

See pricing for current credit cost.

Qwen Image Edit

Image editing via Jobs API. Pass a prompt and source image (base64 or URL fields accepted by Playground / Market).

curl -X POST "https://modkie.com/api/v1/jobs/createTask" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "modkie/qwen-image-edit",
    "input": {
      "prompt": "Replace the background with a soft studio gray",
      "image_url": "https://example.com/product.jpg",
      "width": 1024,
      "height": 1024
    }
  }'

Then poll recordInfo the same way as WAN.

Qwythos 9B

Qwythos is a chat model on /qwythos-9b, not the async Jobs API.

Shared API

POST /api/gateway/chat/completions

{
  "provider": "runpod_vllm",
  "model": "modkie/qwythos-9b",
  "messages": [{ "role": "user", "content": "Hello" }]
}

Dedicated Pod

1. POST /api/runpod/deployments/ensure with { "marketModelSlug": "qwythos-9b-claude-mythos-5-1m" }
2. Poll GET /api/runpod/deployments/{id} until status: ready
3. Call chat with provider: runpod_dedicated_openai and the returned gatewayModelId

Deploying a Pod does not charge credits; runtime billing starts when status is ready. See also Chat for the OpenAI-compatible gateway shape.

Next steps

• Market Quickstart — first call walkthrough
• Payload examples — commercial catalog models
• createTask · recordInfo · Webhooks