Update CHANGES.md

CHANGES.md

@@ -68,6 +68,7 @@
 
 * In Python, [RunInference](https://beam.apache.org/documentation/sdks/python-machine-learning/#why-use-the-runinference-api) now supports loading many models in the same transform using a [KeyedModelHandler](https://beam.apache.org/documentation/sdks/python-machine-learning/#use-a-keyed-modelhandler) ([#27628](https://github.com/apache/beam/issues/27628)).
 * In Python, the [VertexAIModelHandlerJSON](https://beam.apache.org/releases/pydoc/current/apache_beam.ml.inference.vertex_ai_inference.html#apache_beam.ml.inference.vertex_ai_inference.VertexAIModelHandlerJSON) now supports passing in inference_args. These will be passed through to the Vertex endpoint as parameters.
+* Added support to run `mypy` on user pipelines ([#27906](https://github.com/apache/beam/issues/27906))
 
 ## Breaking Changes
 
@@ -127,7 +128,6 @@
   * Transform service controller container.
   * Spark3 job server container.
 * Added support for batched writes to AWS SQS for improved throughput (Java, AWS 2).([#21429](https://github.com/apache/beam/issues/21429))
-* Added support to run `mypy` on user pipelines ([#27906](https://github.com/apache/beam/issues/27906))
 
 ## Breaking Changes