Playground

The Playground is the zero-code half of Grafilab Models. Open Models Playground to browse the live catalogue, run a prompt, generate images and video, and compare models — billed from the same credits and API key you use from code.

The catalogue is split into three category tabs, each showing how many models it holds: Text, Image, and Video.

• Search models filters the grid by name; Clear resets it.
• Sort by offers Price: low to high, Price: high to low, Context: high to low, Name: A to Z, and Newly added.
• The How this sort works button next to it explains the price ordering: Ranked on a blended rate: 3 parts input to 1 part output, the usual usage mix. Neither price column on its own will read in order.
• What’s New? highlights recent additions to the catalogue.

Each text card shows Context, Input, Cached, and Output — the context window and the three per-token rates. Image cards read from MYR 0.05 per generation, video cards from MYR 0.20 per second. A card whose model is not open to you carries a ribbon instead of the usual Try Model button: Coming Soon (with Not yet available) or By Invitation (with Invitation required).

For the exact string to send as model, read the Model ID column on the API page, or call GET /v1/models.

Click Try Model on a text card to open the chat view. The right-hand Model Configuration panel controls the call:

• API Key — required. Pick one of your keys; the run is billed and attributed to it. If the select reads No API keys available, create one first — see API Keys.
• Pricing — the input and output rate for the selected model, per charge unit.
• Temperature — randomness, from Precise to Creative.
• Top P — diversity via nucleus sampling.
• AI Insights — extra analysis alongside the answer.

Type a prompt and send it. The answer streams into the transcript, and the same prompt, model id and key behave the same way through the API.

The Image and Video tabs swap the chat box for generation parameters. Which ones appear depends on the model:

• Aspect Ratio — Square 1:1, Portrait 9:16, Cinema 16:9, 4:3, 3:4, 21:9.
• Resolution.
• Number of images.
• Negative Prompt — what to keep out of the result.
• Image Guidance / Omni Reference, Start frame, and Reference images — for models that accept an image as input, including image-to-video.

Some image and video models carry a free quota shown on the card — Free 100 images per model or Free 10 videos per model — which is spent before your credits are.

To run the same prompt against several models at once, use Compare models in the toolbar — see Compare Models.

Playground runs are real API calls against your balance. Two behaviours surprise people:

Your Usage in the toolbar opens Inference API Usage, which totals Playground, comparison, and API calls together — see Inference Usage.

Everything you set here maps straight onto a request from code:

In the Playground	In your code
The model id on the card	model
The key picked under API Key	the api_key you give the SDK
Temperature / Top P	temperature / top_p

Same call from a terminal: Quickstart: your first API call.