AI Models

The AI Models page is puq.ai’s catalog of every model available through the platform. Use it to see what each model does, compare prices in the same billing unit, copy a model’s ID, and try it live in a built-in playground — all without writing code.

Open it from the sidebar (All Models) at /router/models.

Browsing the Catalog

The catalog toolbar lets you narrow down and sort the list:

Search — matches a model’s name, description, ID, author, task type, and tags.
Author — filter by the model’s provider (for example OpenAI, Anthropic). The picker is searchable and shows how many models match each author.
Price — filter by billing unit (see Pricing below), by input or output rate, and by a min/max price range. Prices only compare within the same billing unit, so the filter narrows to one unit at a time.
Sort — Newest first, Name (A–Z/Z–A), Input/Output price (low↔high), or Largest/Smallest context window.
View — switch between Card view and Table view (table view is available on wider screens only).
Type tabs — a row of tabs under the toolbar filters by task type: Text generation, Image generation, Video generation, Image to video, Text to speech, Speech to text, Music generation, Decisions, and Uncategorized. Each tab shows a live count.

Filters, sort, and view are kept in the page URL, so a filtered link can be shared or bookmarked. The list paginates 50 models per page.

Each card or row shows the model’s icon, name, author, task type, a short description, its price (see below), its context window (for text models), and its model ID with a Copy ID button — the ID you pass as model in API calls.

Model Details Page

Click a model to open its details page. The header shows the model’s icon, name, author, task badge, description, and a few facts: Context window (if applicable), Request body format (JSON or multipart), Last updated, Documentation link (if the provider publishes one), and capability tags (for example Tool calling).

A side panel shows the model’s ID (with Copy) and its Pricing, and a Try in playground button that jumps straight to the Playground tab.

The page has three tabs:

Overview

Shows a worked example when the model has one: an example Request (copyable cURL snippet) next to its Response. The response is rendered as a Preview — images in a gallery, audio with a waveform player, video with a player, text as rendered Markdown, or a structured view for Decisions models — with a Raw JSON tab alongside it for the unprocessed response body. If no example is provided, the tab tells you to try the playground instead.

API reference

Lists the model’s documented Input and Output fields (segmented tabs, each showing a field count). Some models define multiple request/response shapes (Variants, for example different endpoints or modes) — pick a variant to see its fields. This tab only appears when the model has documented fields.

This page documents the app experience only. For the HTTP request/response format itself, see the API documentation.

Playground

Lets you run the model directly from your browser using one of your own API keys:

API key — the field is pre-filled with your most recently created key if you have one; otherwise paste a key or click Create key to generate one without leaving the page.
Variant — if the model has more than one input shape, pick which one to run.
Input fields — generated from the model’s schema. Required fields with no default (and any field the example fills in) appear first under Input; everything else is grouped under a collapsible Settings section. Object/array fields are edited as JSON and validated before you can run.
Run — sends the request (Ctrl/Cmd+Enter also works) and shows a live elapsed timer, then the response status and timing.
Output — a Preview tab renders the result (image gallery with full-size zoom, waveform audio player, video player, rendered Markdown with GitHub-flavored formatting and LaTeX math, or the structured Decisions view) alongside a Raw tab with the full JSON response. If the request fails, the status and error message from the API are shown instead.

Playground runs are real API requests, sent straight to puq.ai’s public API with your API key — the same request your code would make. They are metered and billed to your account exactly like any other API call; nothing about playground usage is free or simulated.

Pricing

Prices are shown per billing unit, and a model is always priced in exactly one unit:

Unit	Shown as
Tokens	per 1M tokens
Image	per image
Megapixel	per megapixel
Tile / tile step	per tile / per tile step
Second	per second of video
Character	per 1K characters
Minute	per minute of audio
Request	per request

Models with separate input/output rates (most text models) show both; single-rate models (most image, audio, and video models) show one price labeled with its unit. A video model with per-resolution tiers shows its cheapest tier prefixed with “from.” A price of $0.00 displays as Free; pricing not yet configured for a model shows Pricing isn’t listed for this model yet.

Because prices only compare meaningfully within the same unit, the price filter and the price-based sort options always operate within one billing unit at a time.

Decisions Models

Models of type Decisions return structured answers instead of free text. Their output (in the Overview example, the playground preview, or the model card) is rendered per answer as a probability, a chosen option with confidence, or a score with ranked options — plus the token usage for that call — rather than as plain text or an image.

Best Practices

Filter by Price before comparing models with different billing units — a raw sort mixes per-token and per-image prices, which aren’t comparable.
Use Copy ID on a card, or the model page, to get the exact string to pass as model in your code.
Try a model in the Playground before wiring it into a workflow or integration, so you can see its real output shape and timing first.
Remember playground runs spend from the same balance as your production API usage.