Discover every generative model on our platform...Browse models
Model Market

Research

Notes on how Model Market presents models and runs — product methodology, not a dump of generated papers.

What a model card must show

Every public model lists task type, example output where available, unit price, and activity signals such as runs. The goal is comparison before spend — not a marketing splash.

Price before you run

Unit pricing is shown on the card and again in Predictions. Credits only move when a chargeable job completes. That keeps browser tests and API jobs on the same rules.

Same job in the browser and the API

A Predictions run uses the model schema you will call from code. We treat that parity as a product requirement: fewer surprises when you promote a prompt into production.

What we are not claiming

We do not invent third-party benchmarks or latency leaderboards without a published method. As we publish numbers, they will live here with clear definitions.

Explore the product

See the catalog in practice, or read how credits and keys fit together on the blog.