Choose a model

Use the interactive model catalog to browse models. Choose by the task, accepted inputs, output constraints, and cost. The number of models changes; inspect the live catalog for current availability.

Task	Starting point	Check before integrating
Language or vision-language inference	Vision and reasoning	Supported model IDs, image support, context, streaming, pricing
Image generation or editing	Image inference	Reference images, aspect ratio, resolution, quality, output format
Video generation	Video inference	Start/end frames, duration, resolution, audio, latency
Speech generation	Audio inference	Available voices, text length, language and speech controls

Inspect a model’s contract

For chat models, read GET /v3/models/{model} and inspect chat.input_modalities, chat.max_images_per_request, input_schema, and chat.pricing. The per-model openapi.json route below describes asynchronous media models.

curl --fail-with-body https://api.hedra.com/v3/models \
  -H "Authorization: Key $HEDRA_API_KEY"
curl --fail-with-body https://api.hedra.com/v3/models/gpt-image-2 \
  -H "Authorization: Key $HEDRA_API_KEY"
curl --fail-with-body https://api.hedra.com/v3/models/gpt-image-2/openapi.json \
  -H "Authorization: Key $HEDRA_API_KEY"

Model detail includes input_schema and output_schema. Read its required fields, enums, defaults, file limits, and duration limits. A field accepted by one model may be rejected by another. Use next_cursor when an endpoint returns paginated results. Estimate a media request with POST /v3/models/{model}/estimate using its exact input. Compare quality and cost on representative inputs before switching a production workload to a different model or tier. Keep production model IDs explicit and validate model changes rather than automatically selecting the newest catalog entry.

Reference and model availability

Open API Reference → Run a model for the contract of a particular model. The reference is regenerated from the v3 schema. For models newer than your installed SDK, use its generic submit method with the live model schema. Bring-your-own-model deployments have a different onboarding process from calling an existing catalog model. See Hedra inference.

Start here

Managed inference

Production operations

SDKs and tools

Language API reference

API Reference

Resources

Choose a model

Inspect a model’s contract

Reference and model availability

​Inspect a model’s contract

​Reference and model availability

Inspect a model’s contract

Reference and model availability