Models and Providers

Genesis aggregates models from multiple providers, but applications call them with a Genesis model ID. Do not substitute a vendor display name, marketplace card title, or application-defined alias for the model value.

List models available to a key

curl "$GENESIS_BASE_URL/models" \
  -H "Authorization: Bearer $GENESIS_API_KEY"

GET /models has no request body. Use data[].id as the model field in other endpoints. Repository references also describe type, capability tags, status, and pricing metadata, but optional fields can vary by deployment. Treat id as the primary field and make the client tolerant of unknown metadata.

The result reflects the current key’s access, not every model on the platform. If a model is absent, check:

  1. whether the model is currently available in Model Marketplace;

  2. whether it belongs to the key’s allowed model scope;

  3. whether the key is active, unexpired, and within quota;

  4. whether the model supports the intended endpoint.

Choose a model

Filter by capability first, then compare context, accepted inputs, pricing, and availability in the model details.

Capability

Call entry point

Conversation or vision understanding

/chat/completions, /responses, or /messages

Text or multimodal vectors

/embeddings

Text or multimodal reranking

/rerank

Because the catalog changes, this documentation does not maintain a static table of model IDs. A production application that discovers models dynamically can refresh /models and maintain an explicit fallback mapping. Do not assume that two models accept identical parameters merely because they share a provider or category.

Provider boundaries

Genesis provides a unified entry point, authentication, access policy, metering, and logs. The underlying provider can still affect supported parameters, context limits, streaming details, and errors. Therefore:

  • send only optional parameters documented for the target model;

  • avoid provider-private fields unless the current model details require them;

  • repeat quality, latency, and cost evaluation after changing a model or provider;

  • record the effective model, request ID, usage, and latency in telemetry, but never log keys or sensitive prompts.

For marketplace and debugging procedures, see the Genesis overview. For common endpoint fields, see the Model API reference.