{
"id": "YOUR_MODEL_ID",
"object": "model",
"created": 1626777600,
"owned_by": "example-provider"
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}API reference
Retrieve a model
Read model metadata and detect errors inside HTTP 200.
{
"id": "YOUR_MODEL_ID",
"object": "model",
"created": 1626777600,
"owned_by": "example-provider"
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"param": "<string>",
"code": "<unknown>"
}
}GET /v1/models/{model}
Use Bearer authentication and pass the exact model ID as a URL-encoded path parameter, with no body. Do not mix in Claude-specific authentication headers, which can select a native model response format.
Success returns one model object, without a list’s data wrapper. For an unavailable or nonexistent model, the current implementation may return HTTP 200 with error.code: model_not_found. Check error before id / object. Authentication, gateway, or proxy failures can still use non-200 status codes.
created is a metadata timestamp, not necessarily a publication date. supported_endpoint_types provides deployment metadata hints, not guarantees for specific parameters, streaming, or tools. Retrieval and listing do not have identical filtering rules; neither replaces an actual inference check.
For IDs containing /, a proxy or router may still split the path after %2F encoding. Use the model list and match the complete ID in the client instead of shortening the model name.
Runnable model lookup · Error handling
阅读中文Authorizations
模型调用 API Key,不是控制台登录凭据 / Model API key, not console login credentials
Path Parameters
经过 URL 编码的完整标识 / Complete URL-encoded identifier
Response
成功结构仍需检查业务字段 / Check business fields even on HTTP success
- Option 1
- Option 2
模型列表返回的原始 ID;需要支持当前接口 / Exact ID from the model list; must support this endpoint
Example:
"YOUR_MODEL_ID"
Example:
"model"
元数据时间戳,不代表实际上架时间 / Metadata timestamp, not necessarily the publication date
部署返回的模型归属标识 / Model ownership label returned by the deployment
部署元数据提示,不保证每项具体参数能力 / Deployment metadata hints, not guarantees for every parameter