Models API

List the models your Antix gateway can serve, and look up a single model. The endpoints are drop-in compatible with both the Anthropic and OpenAI Models APIs, so existing SDKs work unchanged — just point their base URL at Antix.

GET /v1/models — list available models
GET /v1/models/{model_id} — retrieve a single model

The id returned here is exactly the string you pass as the model field when calling the inference endpoints (/v1/messages, /v1/chat/completions). See Routing & BYOK for the full list of supported endpoints.

Quick start

# OpenAI-style listing (default)
curl -s "https://antix.antigma.ai/v1/models"

# Anthropic-style listing (note the anthropic-version header)
curl -s "https://antix.antigma.ai/v1/models" -H "anthropic-version: 2023-06-01"

# Retrieve one model
curl -s "https://antix.antigma.ai/v1/models/claude-sonnet-4-6" -H "anthropic-version: 2023-06-01"

About the URL and the version header

Base URL — the catalog lives at the gateway root (https://antix.antigma.ai), so unlike inference traffic it needs no endpoint URL UUID and no Virtual Key. The examples on this page use the bare host on purpose; for actual /v1/messages or /v1/chat/completions calls, point your SDK at your endpoint URL instead.
anthropic-version — 2023-06-01 is the current stable Anthropic API version that the official SDKs send by default. These public endpoints only check whether the header is present, not its value, so any version string (or x-api-key) selects the Anthropic shape just as well.

Authentication

These endpoints are public — no API key or endpoint URL is required to call them. They return only the catalog of model names the gateway can serve; they never expose credentials, pricing, or routing internals.

note

The model id returned here is the value you pass as the model field to the inference endpoints (/v1/messages, /v1/chat/completions), which do require a Virtual Key or BYOK credential.

Response format & content negotiation

/v1/models serves two response shapes from the same URL, chosen by the request headers:

If the request includes…	You get the…
`anthropic-version` header (regardless of other headers)	Anthropic shape
`x-api-key` and no `Authorization` header	Anthropic shape
neither of the above (e.g. only `Authorization: Bearer …`, `x-api-key` alongside `Authorization`, or no auth header)	OpenAI shape

This mirrors how the official SDKs send requests, so:

the Anthropic SDK (which always sends anthropic-version) automatically receives the Anthropic shape, and
the OpenAI SDK automatically receives the OpenAI shape.

You can force a shape from curl by adding or omitting the anthropic-version header. (Header presence is what matters — the value is not validated by these public endpoints.)

Which models are returned

A model appears only if all three of the following are true:

Configured — it is defined in the gateway's model configuration.
Credentialed — the platform API key for its provider is loaded at startup. For example, if the gateway starts without XAI_API_KEY, no xAI models are listed.
Priced — it has an active price entry in the gateway's rate card.

The price list is read live (with a short cache, ~5 minutes), so when pricing is refreshed from upstream sources the listing updates without a redeploy.

Ordering. Models are returned sorted alphabetically by id.

owned_by / providers. Depending on which provider keys are loaded, the provider (the owned_by field in the OpenAI shape) is one of: anthropic, openai, google, alibaba, deepseek, zai, xai.

Pagination. The full supported catalog is returned in a single page. There is no cursor pagination — the Anthropic-shape has_more is always false, and limit / before_id / after_id query parameters are not used. (The Anthropic list envelope is still returned in full so the Anthropic SDK's auto-paginator works correctly over the one page.)

`GET /v1/models` — List models

List the models the gateway can serve.

Request


Method	`GET`
Path	`/v1/models`
Auth	none

Headers

Header	Required	Notes
`anthropic-version`	no	Presence selects the Anthropic response shape. Conventional value: `2023-06-01`.
`x-api-key`	no	Also selects the Anthropic shape, but only when `Authorization` is not also present on the request — if both are sent, the OpenAI shape wins.

Query parameters

Param	Type	Default	Notes
`return_wildcard_routes`	boolean	`false`	OpenAI shape only. When truthy (`True`/`true`/`1`/`yes`), returns one `"{provider}/*"` wildcard entry per provider instead of one entry per model. Used by LiteLLM-style clients.

Response — Anthropic shape

200 OK

{
  "data": [
    {
      "id": "claude-sonnet-4-6",
      "type": "model",
      "display_name": "Claude Sonnet 4 6",
      "created_at": "1970-01-01T00:00:00Z"
    }
  ],
  "has_more": false,
  "first_id": "claude-sonnet-4-6",
  "last_id": "claude-sonnet-4-6"
}

Field	Type	Description
`data`	array	The list of model objects.
`has_more`	boolean	Always `false` — the full catalog is returned in one page.
`first_id`	string \| null	`id` of the first item in `data`, or `null` if empty.
`last_id`	string \| null	`id` of the last item in `data`, or `null` if empty.

Anthropic model object

Field	Type	Description
`id`	string	The model identifier — pass this as the `model` field to the inference API.
`type`	string	Always `"model"`.
`display_name`	string	Human-readable label derived from the id.
`created_at`	string (RFC 3339)	Release timestamp. Antix does not track release dates, so this is the epoch fallback `"1970-01-01T00:00:00Z"`.

Response — OpenAI shape

200 OK

{
  "object": "list",
  "data": [
    {
      "id": "claude-sonnet-4-6",
      "object": "model",
      "created": 0,
      "owned_by": "anthropic"
    }
  ]
}

Field	Type	Description
`object`	string	Always `"list"`.
`data`	array	The list of model objects.

OpenAI model object

Field	Type	Description
`id`	string	The model identifier — pass this as the `model` field to the inference API.
`object`	string	Always `"model"`.
`created`	integer	Unix timestamp. Antix does not track release dates, so this is `0`.
`owned_by`	string	The provider that owns the model (e.g. `anthropic`, `openai`).

With ?return_wildcard_routes=True:

{
  "object": "list",
  "data": [
    { "id": "anthropic/*", "object": "model", "created": 0, "owned_by": "anthropic" },
    { "id": "openai/*",    "object": "model", "created": 0, "owned_by": "openai" }
  ]
}

Examples

curl — OpenAI shape

curl -s "https://antix.antigma.ai/v1/models"

curl — Anthropic shape

curl -s "https://antix.antigma.ai/v1/models" -H "anthropic-version: 2023-06-01"

Anthropic Python SDK

import anthropic

client = anthropic.Anthropic(
    base_url="https://antix.antigma.ai",
    api_key="sk-antix-<your-key>",  # not validated by this endpoint
)

for model in client.models.list():
    print(model.id, "-", model.display_name)

OpenAI Python SDK

from openai import OpenAI

client = OpenAI(
    base_url="https://antix.antigma.ai/v1",
    api_key="sk-antix-<your-key>",
)

for model in client.models.list().data:
    print(model.id, "-", model.owned_by)

`GET /v1/models/{model_id}` — Retrieve a model

Fetch a single model by its id.

Request


Method	`GET`
Path	`/v1/models/{model_id}`
Auth	none

Path parameters

Param	Type	Description
`model_id`	string	The model id, e.g. `claude-sonnet-4-6`. Must be a single path segment (ids never contain `/`).

The same content-negotiation rule applies: send anthropic-version (or x-api-key) for the Anthropic shape, otherwise the OpenAI shape.

Response

200 OK — a single model object in the negotiated shape (no list envelope).

Anthropic shape

{
  "id": "claude-sonnet-4-6",
  "type": "model",
  "display_name": "Claude Sonnet 4 6",
  "created_at": "1970-01-01T00:00:00Z"
}

OpenAI shape

{
  "id": "claude-sonnet-4-6",
  "object": "model",
  "created": 0,
  "owned_by": "anthropic"
}

The object fields are identical to the per-model objects documented under List models.

Errors

A model that is not served — unknown, its provider key is not loaded, or it is unpriced — returns 404 Not Found in the negotiated error shape. (This is intentional: the endpoint does not distinguish "not configured" from "not currently serveable", so it never reveals that a model merely exists in config.)

Anthropic shape

{
  "type": "error",
  "error": {
    "type": "not_found_error",
    "message": "model: grok-3"
  }
}

OpenAI shape

{
  "error": {
    "message": "The model `does-not-exist` does not exist or you do not have access to it.",
    "type": "invalid_request_error",
    "code": "model_not_found"
  }
}

Examples

curl

# Anthropic shape
curl -s "https://antix.antigma.ai/v1/models/claude-sonnet-4-6" -H "anthropic-version: 2023-06-01"

# OpenAI shape
curl -s "https://antix.antigma.ai/v1/models/claude-sonnet-4-6"

Anthropic Python SDK

model = client.models.retrieve("claude-sonnet-4-6")
print(model.id, model.display_name)

OpenAI Python SDK

model = client.models.retrieve("claude-sonnet-4-6")
print(model.id, model.owned_by)

Status & error reference

Status	When	Body
`200 OK`	List request (always), or retrieve of a served model	List envelope / single model object
`404 Not Found`	Retrieve of a model that is not served	Negotiated error envelope (see above)

The list endpoint always returns 200; if nothing is serveable the data array is empty (has_more: false, first_id/last_id null in the Anthropic shape). For the full error-code reference across all endpoints, see Error Handling.

Notes & FAQ

Why is a model I expected missing? It must be (1) in the gateway's model configuration, (2) backed by a loaded provider key, and (3) priced. A model configured but missing any of these will not appear, and retrieving it returns 404.

Why is created_at the epoch (1970-01-01) / created 0? Antix does not track per-model release dates. The Anthropic API explicitly sanctions an epoch fallback when the release date is unknown.

Why is display_name only in the Anthropic shape? The OpenAI model object has no display_name field; we omit it there to stay faithful to that API. The Anthropic value is derived from the id (Antix has no curated display names).

Does ordering match Anthropic's API? Anthropic orders by release date (newest first). Antix has no release dates, so it orders alphabetically by id.

How fresh is the list? Pricing is read through a ~5-minute cache, so changes from upstream pricing imports appear within a few minutes without a redeploy. The set of configured models and loaded provider keys is fixed at gateway startup.

Quick start​

Authentication​

Response format & content negotiation​

Which models are returned​

GET /v1/models — List models​

Request​

Response — Anthropic shape​

Anthropic model object​

Response — OpenAI shape​

OpenAI model object​

Examples​

GET /v1/models/{model_id} — Retrieve a model​

Request​

Response​

Errors​

Examples​

Status & error reference​

Notes & FAQ​

Quick start

Authentication

Response format & content negotiation

Which models are returned

`GET /v1/models` — List models

Request

Response — Anthropic shape

Anthropic model object

Response — OpenAI shape

OpenAI model object

Examples

`GET /v1/models/{model_id}` — Retrieve a model

Request

Response

Errors

Examples

Status & error reference

Notes & FAQ