Sort of. It used to be $0.10 per 1000 messages. Now it's $0.01 per 10000 requests. Just how much cheaper this turns out to be depends on your ratio of messages to requests.
Real world example: Your job is to send out a few million messages somehow (SMS perhaps). You want to have 100 worker nodes sending out the messages. You populate your queue service with the messages you want to send, then the worker nodes ask the queue service for the next message to send until no more work remains.
Like most Amazon hosted services, they are providing a service so you don't have to host your own hardware and software. You can run your own queue service using RabbitMQ, JMS, or other distributed work queues.
I've been thinking you could make a cheap twitter clone rather easily.
Each active client that checks for short messages every 10 seconds will cost you about one cent per day if there are no messages.
If there are messages, it will cost you 18 cents per gigabyte. Assuming your messages average 128 bytes, every 500k or so messages will cost you another penny.
Of course, you'll probably need at least a few EC2 instances to move messages around between the queues. Not counting your development time, that would probably end up being the largest expense.
