Skip to content
MGModel Gate
Sign inGet started

Endpoints

Supported Model Gate HTTP endpoints with links to complete request and response examples.

Use the inference host configured for your account. The standard Individual API is https://api.model-gate.com. The following table lists every canonical route and method in the gateway route policy. A documented path still requires the appropriate active model, reviewed contract, identity and admission.

MethodsCanonical pathReference and purpose
POST/v1/chat/completionsChat completion
POST/v1/messagesAnthropic Messages
POST/v1/messages/count_tokensMessages token count
POST/v1/responsesResponses inference
POST/v1/embeddingsEmbedding vectors
POST/v1/images/generationsCreate one idempotent generation intent
POST/v1/images/editsCreate one idempotent multipart edit intent
GET/v1/images/catalogSeparate reviewed route offers
POST/v1/images/quoteNormalize a request and quote its price; no dispatch
GET/v1/images/operationsScoped history or idempotency lookup
GET/v1/images/operations/{id}Operation state, usage and file descriptors
GET/v1/images/operations/{id}/resultSaved compatible JSON result
GET/v1/images/operations/{id}/files/{index}Private checked output bytes
POST/v1/images/operations/{id}/retry-downloadsRetry downloads of an existing supplier response only
GET/v1/modelsList enabled key-specific models
GET/v1/models/{id}Resolve one visible model or alias
GET/v1/balanceAuthorized account balance
GET, HEAD/healthLightweight process health
GET, HEAD/health/liveLiveness
GET, HEAD/health/readyService readiness
GET, POST/v1/filesList files or upload batch JSONL
GET, DELETE/v1/files/{id}Read file metadata or delete a file
GET/v1/files/{id}/contentDownload stored batch JSONL
GET, POST/v1/batchesList or create compatible OpenAI batches
GET/v1/batches/{id}Read one batch
POST/v1/batches/{id}/cancelRequest batch cancellation
GET, POST/v1/messages/batchesList or create Message Batches
GET, DELETE/v1/messages/batches/{id}Read or delete a Message Batch
POST/v1/messages/batches/{id}/cancelRequest Message Batch cancellation
GET/v1/messages/batches/{id}/resultsStream stored Message Batch results

Unknown paths return 404 endpoint_not_supported. Unsupported methods return 405 method_not_allowed and Allow. There is no general /v1/* pass-through. Browser preflight is separate from the application methods listed here; it does not grant permission to call an otherwise unsupported method. Files and Batch resource paths accept their existing trailing-slash aliases; do not append slashes to inference or Images paths. Health supports GET and HEAD.

Images generation/editing use saved operations and model_gate.async, not generic Async or Batch. Catalog and quote do not generate images or reserve money. See Images operations and Playground. Partner routes use a separate host and Partner API contract.

Audio transcription

MethodPathReference
GET/v1/audio/catalogTranscription — Read available audio offers
POST/v1/audio/quoteTranscription — Validate an uploaded recording and quote without transcription
POST/v1/audio/transcriptionsTranscription — Create one saved transcription intent
GET/v1/audio/operationsTranscription — Read scoped audio history or intent lookup
GET/v1/audio/operations/{id}Transcription — Read one transcription operation
GET/v1/audio/operations/{id}/resultTranscription — Read the saved transcription response

Video operation methods

MethodPathPurpose
POST/v1/videosCreate one video intent
POST/v1/videos/quoteValidate settings and obtain a nonpaid quote
GET/v1/videos/catalogAvailable video offers for a model
GET/v1/videos/operationsScoped history or idempotency lookup
GET/v1/videos/operations/{id}State, price and result manifest
GET/v1/videos/operations/{id}/resultSaved result manifest, not video bytes
GET/v1/videos/operations/{id}/contentPrivate MP4 bytes; Range supported

See Video generation API. These JSON create/quote methods are separate from generic Async/Batch and supplier multipart.