Model API

Cosmos 3 Super Image

Cosmos 3 Super from NVIDIA creates detailed images from text prompts, with optional prompt expansion and an agentic mode that refines the result over several passes.

Model id

Use this in every request

nvidia/cosmos-3-super/text-to-image

Run it

Quote the input, start the generation with the quote, then poll the job. Set RAYWAKE_API_KEY first; see Authentication.

# 1. Price the exact input (quotes are valid for up to 10 minutes)
curl -s https://api.raywake.com/v1/quotes \
  -H "Authorization: Bearer $RAYWAKE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "nvidia/cosmos-3-super/text-to-image",
  "input": {
    "prompt": "A red ceramic cup on a wooden table, soft morning light"
  }
}'

# 2. Start the generation with the quote_id from step 1
curl -s https://api.raywake.com/v1/generate \
  -H "Authorization: Bearer $RAYWAKE_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{"model": "nvidia/cosmos-3-super/text-to-image", "quote_id": "QUOTE_ID", "input": {"prompt":"A red ceramic cup on a wooden table, soft morning light"}}'

# 3. Poll until the status is terminal
curl -s https://api.raywake.com/v1/jobs/JOB_ID \
  -H "Authorization: Bearer $RAYWAKE_API_KEY"

Input

Input fields for nvidia/cosmos-3-super/text-to-image
Name	Type	Description
`prompt`required	string	Text prompt describing the image to generate.
`negative_prompt`	string	Content to steer the generation away from (colors, objects, artifacts).
`enable_prompt_expansion`	boolean	Expand the prompt with OpenRouter before image generation. When enabled, the prompt is rewritten into the dense structured-JSON format Cosmos3 was trained on; generation falls back to the raw prompt if expansion fails. Default: `false`
`enable_agentic_generation`	boolean	Automatically generate and compare multiple candidate images, then refine the prompt between rounds to better match the original request. This can improve prompt adherence but increases latency and billable image generations. Default: `false`
`agentic_max_iterations`	number (enum)	Maximum number of refinement rounds when agentic generation is enabled. Default: `2` One of: `2`
`agentic_samples_per_iteration`	number (enum)	Candidate images to generate and judge per agentic iteration. The best candidate advances to the next rewrite stage. Default: `2` One of: `2`
`agentic_early_stop`	boolean	Stop early when a candidate image is already a strong match for the prompt. Default: `true`
`image_size`	string (enum)	The size of the generated image. Each edge is clamped to 512-1280px (multiples of 16). Default: `square_hd` One of: `square_hd`, `square`, `portrait_4_3`, `portrait_16_9`, `landscape_4_3`, `landscape_16_9`
`num_inference_steps`	number (enum)	Number of denoising steps. More steps yield higher quality but take longer. Default: `28` One of: `28`
`guidance_scale`	number	Classifier-free guidance scale. Higher values increase prompt adherence at the cost of diversity. Default: `4`
`num_images`	integer	The number of images to generate. Default: `1`
`seed`	integer	The same seed and prompt given to the same model version will produce the same image every time.
`sync_mode`	boolean	If `True`, the image is returned as a data URI and the output data won't be available in the request history. Default: `false`
`output_format`	string (enum)	The format of the generated image. Default: `jpeg` One of: `jpeg`, `png`

Output

A succeeded job lists its results under outputs, stored on our side. See Jobs for every field.

Job excerpt

"outputs": [
  { "type": "image", "url": "https://…", "mime_type": "image/png", "thumbnail_url": "https://…" }
]

Pricing

1 image$0.07/ image

The exact charge for your input comes from a quote. How charges settle