faucet
Request access
Guides

Choosing a model

How namespaced slugs and aliases work, and how to ask the gateway what your own key can reach.

Model ids are namespaced provider/model, and that slug is a contract: it is never repointed at a different model, only at a newer snapshot of the same one. Pinning anthropic/claude-opus-5 in your code is safe.

Aliases let a bare name resolve too — claude-opus-5 reaches the same entry — so a codebase migrating off a single provider can change only its base URL and keep working.

Ask what your key can reach

The public Models reference lists every slug Faucet carries. What your key can reach is narrower, because the gateway also applies that key's allowlist and skips providers holding no usable credential:

curl https://api.intfaucet.com/v1/models \
  -H "Authorization: Bearer $FAUCET_API_KEY"

This is the authoritative list

GET /v1/models is the only answer narrowed to what a given key may reach. The reference page is generated at build time and can see neither your allowlist nor which credentials are usable. Prices are in the console, where the figure shown is the figure you are charged.

One model, one endpoint

Faucet serves four kinds of model and each answers on its own path. Sending an embedding model to /v1/chat/completions, or a chat model to /v1/embeddings, or a transcription model anywhere but /v1/audio/transcriptions, is a 400 that names the endpoint you actually wanted — not a 404, because the model does exist and the slug is not what needs fixing. The message is derived from the model you named rather than from the endpoint that rejected it, so it always points somewhere real.

KindEndpointGuide
Chat/v1/chat/completionsChat
Embedding/v1/embeddingsEmbeddings
Image/v1/images/generations, /v1/images/editsImages
Transcription/v1/audio/transcriptionsTranscription

The native Anthropic subset

GET /v1/models covers every Faucet endpoint. The models that additionally answer in Anthropic's native format are the chat models whose owned_by field is anthropic: they work on /v1/messages — and Claude Code's /v1/messages/count_tokens — as well as on /v1/chat/completions. Other providers currently use the OpenAI-format endpoints only.

The context window is per request

The window is a ceiling on a single request, not on a conversation. Overrun it and the provider's own 400 reaches you with its message intact. Summarize or drop old turns before that point; the window for each model is in the Models reference, and Chat covers what a growing thread costs.