API Reference
The API Reference provides the technical information required to integrate InstantGPU services into your applications.
Authentication #
Authenticate requests using your InstantGPU API key in the Authorization header:
Authorization: Bearer YOUR_API_KEY
API Request #
A typical API request includes:
Endpoint URL
Target model HTTPS gateway URL.
HTTP Method
POST for inference / GET for status.
Authentication
Bearer YOUR_API_KEY header.
Content Type
application/json
Request Parameters
JSON payload containing model input, hyperparameters, and options.
Request Example
curl -X POST "https://api.instantgpu.ai/v1/endpoints/image-generation/runsync" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"prompt": "Generate an image of a futuristic city",
"width": 1024,
"height": 1024,
"num_outputs": 1
}'
Request Parameters #
Parameters depend on the selected endpoint or model. Common parameters include:
Prompt / Input Text
The core instruction or content string supplied to the model.
Image Dimensions
Width and height dimensions for visual models (e.g. 512x512, 1024x1024).
Number of Outputs
How many sample completions, variations, or images to return.
Model Settings & Hyperparameters
Temperature, top_p, max_tokens, guidance_scale, seed, and inference steps.
Response #
The API returns a structured JSON payload with the output generated by the selected service:
{
"id": "req_01jf8b9a2x3y4z",
"status": "COMPLETED",
"executionTimeMs": 842,
"output": {
"images": [
"https://storage.instantgpu.ai/outputs/futuristic-city.png"
]
}
}
Error Handling #
Applications should handle common HTTP status codes and error bodies gracefully:
401 Unauthorized
Invalid or missing InstantGPU API key. Verify Bearer token.
400 Invalid Request
Missing required parameters or malformed JSON payload.
404 Not Found
Endpoint unavailable or wrong URL path.
429 Rate Limit
Request volume exceeded quota. Back off and retry.
504 Timeout
Execution exceeded maximum worker timeout threshold.
500 Server Error
Internal worker exception. Review runtime logs in dashboard.
Rate Limits #
Rate limits may vary depending on the service, endpoint tier, and account configuration. Applications should handle rate-limit responses gracefully by respecting the Retry-After response header and implementing exponential backoff.