NVIDIA: SANA-Sprint
sana-sprint- Price
- From $0.002621
- Modalities
- Image
About SANA-Sprint
SANA-Sprint is an open-weight text-to-image model from NVIDIA's NVlabs, distilled so it can produce an image in one to four sampling steps instead of the usual twenty or more. It is built on the SANA linear diffusion transformer, uses the Gemma2-2B-IT language model to read prompts, and targets 1024-pixel images. Choose it when throughput and latency matter more than fine detail.
Where it works well
- It needs only 1 to 4 sampling steps, which is the reason it responds in a fraction of the time of standard diffusion models.
- NVIDIA reports 7.59 FID and 0.74 GenEval in a single step, ahead of FLUX-schnell on those two measures.
- A single model handles any step count from 1 to 4, so you can trade latency for quality per request.
- Weights are published under Apache 2.0, which makes it clear for commercial pipelines.
When to choose another model
- The model card says it does not render complex legible text, so posters and UI mockups with copy need another model.
- It is not perfectly photorealistic and can get details like fingers wrong.
- It is text-to-image only; there is no editing or reference input.
Getting started
Create API key
Create a key in Console, then use it with every model on the platform.
Send your first request
Copy the example for your language and run it against the endpoint.
Text to imageTokenLab endpointPOST/v1/images/generationscurl -X POST "https://api.tokenlab.sh/v1/images/generations" \ -H "Authorization: Bearer sk-xxx" \ -H "Content-Type: application/json" \ -d '{ "operation": "text-to-image", "model": "sana-sprint", "prompt": "A minimalist product photo of matte black headphones on a soft blue background." }'
Parameters
Text to image
| Parameter | Values and limits | Default | Notes |
|---|---|---|---|
guidance_scale | ≥ 0, ≤ 20 | Not provided | Not provided |
n | 1 | 1 | Not provided |
negative_prompt | Not provided | Not provided | Not provided |
operation | Not provided | Not provided | Not provided |
output_format | jpeg, png | jpeg | Not provided |
prompt | Not provided | Not provided | Not provided |
seed | Not provided | Not provided | Not provided |
size | Not provided | 1024x1024 | Use WIDTHxHEIGHT. An omitted size uses 1024x1024; exact custom-dimension bounds are not yet verified. |
steps | ≥ 1, ≤ 20 | 2 | Not provided |
Fields outside this contract are rejected.
Pricing
The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.
Text to image
per request- Official price
- $0.002621
- Verified price
- —
- Discount
- —
| Official priceper request | Verified priceper request | Discount | |
|---|---|---|---|
| Text to image | $0.002621 | — | — |
Usage & activity
Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.
Usage & availability
Last 24 hours- Requests
- Success rate
- P95 latency
- Total tokens
Data is based on aggregate user requests, excluding status checks.
Open in Console
Open SANA-Sprint in Console with a prompt ready to edit or send.
Help me create an image with sana-sprint using text-to-image at /v1/images/generations. Show the result, cost, and any limits I should know.
Use cases
Best for- Text to Image
Live previews
Show an image as the user finishes typing a prompt, where a response under a second changes how the tool feels.
High-volume asset generation
Create thousands of backgrounds, icons, or placeholder illustrations where per-image speed drives the job size.
Prompt exploration
Try dozens of wording variants quickly, then rerun the best on a slower, higher-detail model.
Consumer app features
Add avatar or wallpaper generation inside an app that cannot wait through a 20-step sampler.
Prompt examples
A cozy reading nook with a window seat, rainy afternoon, soft lamp light, painterly illustration.
Isometric low-poly island with a lighthouse and tiny boats, pastel colors, clean background.
Macro photo of a dew-covered spider web at sunrise, bokeh background.
FAQ
What is SANA-Sprint exactly?
It is a distilled version of NVIDIA's SANA text-to-image model. Continuous-time consistency distillation lets it sample in one to four steps, giving very fast generation while keeping quality competitive with larger multi-step models.
How fast is it?
NVIDIA reports about 0.1 seconds per 1024-pixel image on an H100 GPU. Hosted latency will be higher because of queueing and network time, but it is designed to be much quicker than a 20-step diffusion model.
What resolution does it produce?
The model card describes it as designed for 1024-pixel images with multi-scale height and width, so request sizes near that range for the most dependable results.
Can it write readable text inside images?
Not reliably. The model card lists complex legible text as a limitation. For signage, posters, or screenshots with copy, use a model known for typography.
What is the license?
The model card states Apache License 2.0, which allows research and commercial use. Attribution terms and any conditions are spelled out on the card itself.
How much does SANA-Sprint cost?
On TokenLab, SANA-Sprint costs $0.002621 per request. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.
Which endpoint should SANA-Sprint use?
Use https://api.tokenlab.sh/v1/images/generations for SANA-Sprint. The request example below shows the matching code shape.
Which operations does SANA-Sprint support?
SANA-Sprint supports Text to image. Select an operation above to see its endpoint and request example.
Compare SANA-Sprint
Sources
Reviewed Oct 2, 2026