Models¶

These are the models you can call. Prices, rate limits and live status are on each model's page in the Console, under Inference > Models.

Model	Provider	Model ID
Kimi K3.1 Pro	Moonshot AI	`kimi-k3-1-pro`

Model IDs¶

Send the model ID as model in your requests. You'll find it on the model's page in the Console, under the model name, with a copy button. Use it exactly as shown.

Enable a model¶

Your project can call a model only after the model is enabled for it. Click Enable on the model's card under Inference > Models, or on the model's page.

If a model shows Out of capacity, it isn't open to new projects right now. Projects that already enabled it keep using it. Try again later or contact support.

Inference Endpoint API¶

Send inference requests to https://inference.verda.com/v1 with an inference API key. These endpoints are supported:

Endpoint	What it does
`POST /v1/chat/completions`	Generates a response with the OpenAI Chat Completions API, with or without streaming
`POST /v1/responses`	Generates a response with the OpenAI Responses API, used by clients such as Codex
`GET /v1/models`	Lists the model IDs your project can call, in the OpenAI format

Other OpenAI endpoints, such as embeddings, images or audio, return 404 with the code endpoint_not_supported. See OpenAI compatibility.

For example, to list the models you can call:

curl "https://inference.verda.com/v1/models" \
  -H "Authorization: Bearer $VERDA_INFERENCE_API_KEY"

Model access management¶

You can also list, enable and disable models with the Verda API. These calls use a Cloud API access token, not an inference API key. See API Credentials.

Endpoint	What it does
`GET /v1/inference/models`	Lists the models available to your project, with prices in your billing currency
`GET /v1/inference/models/{code}`	Returns one model
`GET /v1/inference/subscriptions`	Lists the models enabled for your project
`PUT /v1/inference/subscriptions/{model_code}/standard`	Enables a model
`DELETE /v1/inference/subscriptions/{model_code}/standard`	Disables a model
`GET /v1/inference/subscriptions/status`	Tells you whether inference is active or paused for your project
`GET /v1/inference/usage`	Returns your token usage per day and model

For example, to enable Kimi K3.1 Pro:

curl -X PUT "https://api.verda.com/v1/inference/subscriptions/kimi-k3-1-pro/standard" \
  -H "Authorization: Bearer $VERDA_API_TOKEN"

If the model can't be enabled, the response's error.code says why, for example insufficient_balance when the project balance is too low, or model_out_of_capacity when the model isn't open to new projects right now.

When a model is deprecated¶

When we deprecate a model, we mark it as deprecated in the Console and in the API, with the date it stops being served:

In the Console, the model shows Deprecating on and the date until then, and Deprecated on and the date after it. A deprecated model can't be enabled or used in the Playground.
In the API, GET /v1/inference/models returns the date in deprecation_date and the model to move to in replacement_code.