Google Gemini 2.5 Flash retired in Databricks

The Google Gemini 2.5 Flash model has been discontinued and is no longer available for inference requests in Foundation Model APIs, for both pay-per-token and provisioned throughput modalities.

Use cases

Imagine you have AI applications using Gemini 2.5 Flash to process text and images. With the model retirement, these requests will fail unless updated.

• Identify which endpoints and users still depend on retired models.
• Migrate applications to newer recommended models.

How to use

Run a SQL query against the system catalog to find workloads using the retired model and identify their owners.

SELECT
   eu.requester,
   se.endpoint_name,
   se.entity_name,
   COUNT(*) AS request_count,
   SUM(eu.input_token_count) AS total_input_tokens,
   SUM(eu.output_token_count) AS total_output_tokens,
   MIN(eu.request_time) AS first_request,
   MAX(eu.request_time) AS last_request
 FROM system.serving.endpoint_usage eu
 JOIN system.serving.served_entities se
   ON eu.served_entity_id = se.served_entity_id
 WHERE LOWER(se.entity_name) LIKE '%gemini-2.5-flash%'
 GROUP BY eu.requester, se.endpoint_name, se.entity_name
 ORDER BY request_count DESC

What happens in this code:

1. Joins endpoint usage data with served entities from the system schema.
2. Filters by the name of the retired model.
3. Groups by requester and endpoint name to count requests and token usage.

Limitations and requirements

The model is completely inaccessible. Recommended replacement models are Gemini 3.1 Pro or Gemini 3.5 Flash.