Managed models

Instance version requirements

Managed models require a supported Hologres instance version.

Upgrade your instance from the Hologres console if needed.

Supported regions

Available regions:

• China (Beijing)

• China (Shanghai)

• China (Hangzhou)

• China (Shenzhen)

• Singapore

Purchase model resources

Purchase Hologres AI model resources from the AI model resource purchase page before using managed models.

When purchasing resources, note:

• Pay-as-you-go: Billed based on actual usage with no upfront fees.

• Region selection: Select the same region as your Hologres instance to avoid cross-region charges.

• Billing dimensions: Billing units vary by model: token count, image count, video duration, or audio duration.

Pricing details

For per-model pricing, see Managed models billing.

Model catalog

Hologres managed models support text generation, translation, embedding, image generation/editing, speech-to-text, and video generation. The following table lists the model category, model_type, task type, input/output, notes, and API access for each model.

Deploy a model

After purchasing model resources, deploy managed models from the Hologres console:

1. Log on to the Hologres console and go to the details page of the target instance.

2. In the left-side navigation pane, choose AI Model to open the model management page.

3. Click Deploy Model.

4. Set the model provider to Hologres-managed model.

5. Select the model categories you need. You can deploy multiple models at once.

6. Configure parameters for each model (defaults usually work) and set the retry mechanism.

7. Click Confirm to complete the deployment.

8. After deployment, view deployed models and their status on the model list page.

Model usage

After a model is deployed, you can call it in either of the following ways:

• Call the model through an AI Function: Run SQL statements to call an AI Function. This method suits scenarios in which you process data directly in the warehouse.

• Call the model through the API: Call the AI Function API. This method suits scenarios in which you reuse a deployed model from an application or an AI agent outside Hologres. Only some models support this method. For more information, see the API access column in Model catalog.

SELECT * FROM list_ai_function_infos();

    function_name     |    model_name
----------------------+------------------
 ai_embed             | my_gte_embedding
 ai_classify          | my_qwen32b
 ai_extract           | my_qwen32b

SELECT set_ai_function_info('<FUNCTION_NAME>', '<MODEL_NAME>');

SELECT set_ai_function_info('ai_embed', 'my_gte_embedding');

curl http://hologres-token-cn-hangzhou.hologres.aliyuncs.com:80/anthropic/v1/messages \
  -H "x-api-key: <API_KEY>" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"glm-5.2","max_tokens":1024,"messages":[{"role":"user","content":"Hello"}]}'

SELECT query, extended_cost, *
FROM hologres.hg_query_log
WHERE query LIKE '%ai_gen%'
  AND query_start >= now() - interval '1 hour'
ORDER BY query_start DESC;

{
  "qe_model_usage": {
    "TokenModelUsage": {
      "qwen3.7-max": {
        "tiers": {
          "0-unlimited": {
            "input_tokens": 703,
            "output_tokens": 151,
            "cache_creation_input_tokens": 0,
            "explicit_cache_tokens": 0,
            "implicit_cached_tokens": 0,
            "total_cached_tokens": 0,
            "total_tokens": 854
          }
        },
        "total_tokens": 854
      }
    }
  }
}

1. Purchase model resources

Purchase Hologres AI model resources from the AI model resource purchase page. Billing is pay-as-you-go based on actual usage.

2. Deploy the models

After purchasing resources, open the instance details page in the Hologres console, go to the AI Model tab, select Hologres-managed model as the provider, and deploy the models you need.

This example deploys the following models for the AI Function calls in step 3:

• qwen3.7-plus: Text generation model for inference tasks.

• text-embedding-v4: Embedding model for vector representations.

3. Call AI Functions

After deployment, call models through SQL AI Functions. For the full list and usage details, see AI functions.

Example 1: Text inference (with qwen3.7-plus)

-- Use a managed model for text generation
SELECT ai_gen('qwen3.7-plus', 'Generate an engaging description for the following product: wireless noise-cancelling headphones');

Example 2: Embedding (with text-embedding-v4)

-- Use a managed model to generate embeddings
SELECT ai_embed('text-embedding-v4', 'Hologres is a real-time data warehouse product');