What causes ML_CLASSIFICATION_MODEL_NOT_FOUND?

The classification model instance was dropped manually or as part of a schema cleanup. The code references a model name that was created in a different database or schema. The calling session uses a different current database or schema than where the model was trained. The model was trained in a development schema and a production pipeline references it without a fully qualified name

How do I fix ML_CLASSIFICATION_MODEL_NOT_FOUND?

Step 1: Run SHOW SNOWFLAKE.ML.CLASSIFICATION to list all classification model instances in the current schema.. Step 2: If the model is missing, retrain it using the SNOWFLAKE.ML.CLASSIFICATION constructor with the original training data.. Step 3: Update all prediction calls to use the fully qualified model name (database.schema.model_name) to avoid session-context issues.. Step 4: Verify that the calling role has USAGE privilege on the schema where the model lives.. Step 5: Add a pre-check in the pipeline to confirm the model exists before running predictions.

High severityconfiguration

Power BI Refresh Error:
ML_CLASSIFICATION_MODEL_NOT_FOUND

What does this error mean?

A call to a Snowflake ML Classification model instance failed because the model object does not exist in the specified schema, was dropped, or the caller is using the wrong database or schema context.

Common causes

1The classification model instance was dropped manually or as part of a schema cleanup
2The code references a model name that was created in a different database or schema
3The calling session uses a different current database or schema than where the model was trained
4The model was trained in a development schema and a production pipeline references it without a fully qualified name

How to fix it

1Step 1: Run SHOW SNOWFLAKE.ML.CLASSIFICATION to list all classification model instances in the current schema.
2Step 2: If the model is missing, retrain it using the SNOWFLAKE.ML.CLASSIFICATION constructor with the original training data.
3Step 3: Update all prediction calls to use the fully qualified model name (database.schema.model_name) to avoid session-context issues.
4Step 4: Verify that the calling role has USAGE privilege on the schema where the model lives.
5Step 5: Add a pre-check in the pipeline to confirm the model exists before running predictions.

Frequently asked questions

Does Snowflake ML Classification retrain automatically on new data?

No. The model is a point-in-time snapshot trained on the data you supply. To update it with new data, you must retrain by calling SNOWFLAKE.ML.CLASSIFICATION with the updated training query.

Can I copy a Classification model instance between Snowflake accounts?

No. ML function model instances are local to the Snowflake account and schema where they were trained. Cross-account model sharing is not currently supported for ML Function instances.