Models & naming

Model catalog

The model catalog lists every model offered on this site and can be filtered by vendor, context length, price and input modality. Each model page shows:

Context length and maximum output length; Input / output prices (USD per million tokens), plus cache read and write prices; Supported capabilities such as tool calling, image input and reasoning; Example code.

Model IDs

The model parameter takes a model ID, which usually matches the vendor's official model name:

The catalog and GET /v1/models are the source of truth; the table only illustrates the naming.

Groups and available models

Every API key belongs to a group, which decides:

Which models can be called (a group may have a model allowlist); Which upstream platform the request is sent to; Which prices apply.

Requesting a model the group does not support returns an error, so choose a group that contains the models you need when creating a key. Some groups are composite groups that route to the right vendor based on the model name, so one key can use models from several vendors.

Cross-protocol calls

The same model can be called through different protocols. A Claude model, for example, works with both the Anthropic Messages API and the OpenAI Chat Completions API — the gateway converts the formats for you. Use whichever protocol your code or tool already supports.