Choose an AI model category by the task
Start with the output you need, then check inputs, capabilities and billing terms.
Before choosing a model, write down what you need to deliver: editable text, an image, a video or speech. A model may accept several input types, but our directory groups models by their main output task. Multimodal input does not imply multimodal output.
Next, check the provider’s exact API model ID, supported inputs and outputs, and the mode you need. An image output token rate is usually not the total cost of one image. A per-second video rate may also depend on resolution.
Compare candidates with a fixed task, output specification and platform. Then check the price source and verification date. Provider documentation describes an interface; it does not replace a small test with your own prompts, quality bar and budget. We currently summarize official sources and have not published an independent performance ranking.