Overview

A catalog of hosted chat, embedding, image, and audio models you can try instantly in the Playground — no code required.

AI Inference is a catalog of hosted models — chat, embeddings, image generation, and audio transcription — that you use straight from the console's Playground. Pick a model, give it an input, and see the result on screen.

Why it's useful

Instead of juggling separate tools for chat, search embeddings, image generation, and transcription, the Playground gives you one place to try any of them against the same model catalog. Explore what a model can do, sanity-check a prompt, or generate a one-off result — all without leaving the dashboard.

• Compare models side by side — run the same prompt against a frontier model and a fast-and-cheap model to see how much quality you actually give up, before you pick one for a project.
• Sanity-check a prompt idea — try a system prompt or instruction on a handful of inputs in the Playground before it becomes part of something you rely on regularly.
• Understand cost before committing — every Playground result shows the tokens or units it used, so you can see roughly what a real workload would cost before you build around a model.

Quickstart

1. 1
   Open the Playground

   From the dashboard sidebar, open Playground under Inference.

   

2. 2
   Pick a model

   Choose a model from the catalog — chat, embeddings, image, or audio models are all available.
3. 3
   Try it

   Type a prompt, upload a file, or enter text, then send it and see the result appear in the console.

Comparing models before you decide

Switching the Model field is the fastest way to learn the difference between two models — run the same prompt through a frontier model, then again through a fast-and-cheap model, and compare the replies and the token cost side by side. There's no need to guess which model fits a project; the Playground lets you find out first.

Explore