Moonshot AI Kimi K2.7 model scheduled for retirement

The Moonshot AI Kimi K2.7 model has been scheduled for retirement in Databricks. It will no longer be accessible through Foundation Model APIs starting October 30, 2026.

Use cases

Helps identify active workloads relying on the Kimi K2.7 model so teams can plan their migration before the end-of-life date.

How to use

Run the following SQL query to locate endpoints and users utilizing the deprecated model.

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 '%kimi-k2.7%'
 GROUP BY eu.requester, se.endpoint_name, se.entity_name
 ORDER BY request_count DESC

What happens in this code:

1. Joins endpoint usage telemetry with served entities from the system catalog.
2. Filters by the specific model name slated for retirement.
3. Aggregates request counts by user and endpoint name for auditing purposes.

Limitations and requirements

The Kimi K2.7 model will be fully retired on October 30, 2026. Migration to Kimi K3 must be completed prior to this date to prevent service disruption.

Using it in a pipeline

Update inference configurations and code references from Kimi K2.7 to Kimi K3 across your production AI applications.