Skip to content

Model directory

Every model, side by side.

Browse 88 models from 20 providers across the Chinese and global ecosystems.

Showing 24 of 88 models · sorted by trending

Compare models side by sideUp to four at once on price, context, measured latency and capability — and, for the models AWS Bedrock or Microsoft Foundry also sell, what they charge for the same thing.Open compare →

What is in the catalogue

Every entry records the same fields: input and output price per million tokens, context window, provider, capability category and, where a published result exists, benchmark scores. Recording the same fields for every model is the point — it is what makes two models from different providers actually comparable rather than merely adjacent on a page.

How to read the pricing

Prices are per million tokens and split between input and output, because the two rarely cost the same and the ratio between them decides which model is cheaper for your workload. A summarisation job reads far more than it writes; a generation job does the opposite. A model with cheap input and expensive output can be the economical choice for one and the costly one for the other, so compare against the shape of your own traffic rather than a headline figure.

Context windows

The context window is the total the model can hold at once — your prompt, any retrieved documents, the conversation so far and the reply it is about to write, all counted together. A larger window costs more per request and does not by itself improve an answer, so treat it as a ceiling to design under rather than a score to maximise.

Model directory: compare AI models · Webparam