Skip to main content
POST
Create or Update Model

Authorizations

Authorization
string
header
required

A control-plane management key using the sk-cp- prefix. Inference keys are not accepted by control-plane endpoints.

Body

application/json
model_id
string
required

Caller-facing model name

Required string length: 1 - 255
provider_id
string
required

Provider id the model routes to

Required string length: 1 - 63
Pattern: ^[a-z0-9][a-z0-9_-]*$
input_modalities
enum<string>[]
required

Accepted input modalities

Required array length: 1 - 16 elements
Available options:
text,
image,
audio,
video,
pdf
output_modalities
enum<string>[]
required

Produced output modalities

Required array length: 1 - 16 elements
Available options:
text,
image,
audio,
video,
pdf
upstream_model
string
default:""

Model name sent to the provider, lets model_id be an alias; defaults to model_id

Maximum string length: 255
egress_kind
string | null

Per-model egress adapter override

Required string length: 1 - 63
Pattern: ^[a-z0-9][a-z0-9_]*$
input_price_per_mtok
string
default:0

USD per million input tokens

Pattern: ^\d+(?:\.\d+)?$
output_price_per_mtok
string
default:0

USD per million output tokens

Pattern: ^\d+(?:\.\d+)?$
cache_read_price_per_mtok
string
default:0

USD per million cache-read input tokens

Pattern: ^\d+(?:\.\d+)?$
cache_write_price_per_mtok
string
default:0

USD per million cache-write input tokens

Pattern: ^\d+(?:\.\d+)?$
context_window
integer
default:128000

Context window in tokens

Required range: 1 <= x <= 100000000
max_output_tokens
integer | null

Max completion tokens; requests are clamped to it

Required range: 1 <= x <= 100000000
capabilities
enum<string>[]

Capabilities supported by the model

Maximum array length: 4
Available options:
streaming,
tools,
reasoning,
structured_output
parameter_support
Parameter Support · object

Known support for canonical request parameters; an absent parameter is unknown

Response

Successful Response

data
ModelOut · object
required