Model selection and API access
Compare capabilities and list prices in the public catalog, then use your enterprise API key to confirm access. Choose an endpoint for your workload and check your effective rates in the console.
Find an eligible model ID
GET /api/public/models- Read capabilities, aliases and list rates for technical evaluation. Use models[].id for the model ID, then check access with a contracted enterprise API key.
GET /v1/models- Call with an enterprise API key. data[].id includes model IDs and aliases available under enterprise, department, plan, key, source IP and channel rules.
GET /v1beta/models- A Gemini-shaped list governed by the same access rules. Use it for discovery, then test the intended operation on the target endpoint.
Set BASE_URL and TOKEN_PYXIS_API_KEY as in the quickstart. Check the authenticated list for access and prefer the canonical catalog ID in production configuration.
curl -fsS "https://tokenpyxis.com/api/public/models"
curl -fsS "https://tokenpyxis.com/v1/models" -H "Authorization: Bearer $TOKEN_PYXIS_API_KEY"
curl -fsS "https://tokenpyxis.com/v1beta/models" -H "x-goog-api-key: $TOKEN_PYXIS_API_KEY"Browse the model catalog and list ratesChoose an API protocol
Check the matching capability flag in the public catalog, then verify an actual request with your API key and target model. Upstream capabilities also affect tools, caching and specific parameters.
| Protocol and endpoint | Flag | Use case |
|---|---|---|
Chat CompletionsPOST /v1/chat/completions | chat | New text applications and existing chat.completions workflows. |
ResponsesPOST /v1/responses | responses | Existing Responses requests and clients. Test streaming, tools and multi-turn behavior for your use case. |
Responses CompactPOST /v1/responses/compact | responses_compact | Native context compaction with a non-streaming result. |
MessagesPOST /v1/messages | messages | Apps that already use the Anthropic Messages payload. |
GeminiPOST /v1beta/models/{model}:generateContent | chat | Apps using the contents payload. Safety settings, tools, media and explicit caching require a native Gemini route. The streaming endpoint sends incremental SSE events. |
Images, speech, embeddings, rerank, video and realtime each have their own endpoints. See all endpoints · Media generation and speech recognition
Check rates and usage
A token is a unit of model input or output. Token-priced models typically list input, output and applicable cache rates per million tokens. Image, speech and video endpoints may charge per image, second or call. Check the unit for your model.
Use the public catalog to compare list rates. Base production estimates on your signed agreement. Check effective rates, wallet and plan coverage in the console, then reconcile a sample by request ID.
