Kafka streams single big transformer vs several small ones

I have a topology organized with some chained steps with different responsibilities and I’ve been suggested to combine the transformers for these steps into a single one “for efficiency”.

We can take as an example the following topology sequence:

...
  .transform(SomeFilter::new) // Filter using a transformer to access the context
  .transform(SomeParser::new)       
  .flatTransform(SomeSplitter::new)
...

Do I gain anything by combining these three transformers into a single one?
Which is the recommended approach?