If I make 6 replicas of the service with 3 partitions, will it help speed up reading?

The number of service replicas and the number of partitions are different aspects of scaling. If you have 3 partitions, data is distributed across them. Creating 6 service replicas can help improve availability and request processing parallelism, but it does not speed up reading from partitions by itself if requests are not evenly distributed among replicas and partitions.

To speed up reading, it is important that requests are effectively sharded across partitions, and replicas can serve requests in parallel. Ultimately, 6 replicas with 3 partitions can help if the load is distributed correctly, but simply increasing the number of replicas does not guarantee faster reading.