The Google Cloud Platform, Google’s public cloud infrastructure that rents out compute and storage to run your applications, announced today that the Google Compute Engine’s Custom Machine Types feature is now generally available. This allows customers to fine-tune instance sizes to the exact number of cores and amount of memory they want.

Google first introduced Custom Machine Types in beta in September.

“Since our Beta launch, we have seen customers create virtual machines with novel vCPU and memory ratios that aren’t available from any of the major cloud providers,” Google Cloud Platform’s Sami Iqram wrote in a blog post. “As a result, our customers have saved an average of 19 percent — and as much as 50 percent — on top of our already market-leading prices.”

The feature helps distinguish Google from public cloud market leader Amazon Web Services (AWS), as well as from Microsoft Azure, IBM SoftLayer, and others. At its scale, Google can afford to procure a whole lot of different flavors of hardware — and then add smart software on top. But you could say the same thing about Amazon and Microsoft, and it might not be long before similar functionality becomes more widely available. (As I mentioned at the time of the feature’s beta launch, a smaller public cloud called CloudSigma has had this capability for years.)

Igram added that as of today, Windows and Red Hat Enterprise Linux (RHEL) operating systems can work with Custom Machine Types, alongside CentOS, CoreOS, Debian, OpenSUSE, and Canonical’s Ubuntu.

With, for example, Ubuntu, an instance with minimal CPU cores and memory can cost as little as $4.49 per month out of Google’s U.S. East and Central data center regions ($5 in Asia and Europe) and as much as $1,030 per month with both sliders tuned up to the max at 32 vCPU and 208GB of memory ($1,136.86 in Asia and Europe). Prices are higher for Windows and RHEL.