> ## Documentation Index
> Fetch the complete documentation index at: https://planetscale.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Supported versions

> When creating a PlanetScale Postgres cluster, you can choose from any of the Postgres versions we support.

This page covers Postgres.

PlanetScale Postgres supports versions 17 and 18 of Postgres.

Specifically, the following versions are currently supported:

* 17.12
* 18.7

New databases will be created using the latest version by default and we recommend sticking to that default, but you can choose an older version if you need to.

## Major version upgrades

PlanetScale doesn't currently offer in-place upgrades between major versions of Postgres. To upgrade from Postgres 17 to 18, create a new Postgres 18 database and perform an online migration from your existing PlanetScale Postgres 17 database using our [import guides](https://planetscale.com/docs/postgres/imports/postgres-imports.md).

The same import process also applies if you're migrating from another provider such as RDS, Aurora, or a self-hosted Postgres instance. You can import into any supported major version regardless of where the source database is hosted.

## Benchmarking Postgres 17 vs 18

Our [Benchmarking Postgres 17 vs 18](https://planetscale.com/blog/benchmarking-postgres-17-vs-18) blog post covers sysbench `oltp_read_only` results across multiple EC2 instance types and I/O configurations.

* **Postgres 18's new `io_method=worker` default delivers the best overall read performance**, outperforming both `io_uring` and the legacy synchronous I/O in most scenarios.
* **Local NVMe disks dramatically outperform network-attached storage** (gp3, io2) regardless of Postgres version or I/O setting.

## What's new in Postgres 18

Postgres 18 builds on Postgres 17 with new features, changed defaults, and some deprecations. For full details, see the official [Postgres 18 release notes](https://www.postgresql.org/docs/18/release-18.html).

| Area | What Postgres 18 adds |
| :- | :- |
| **I/O architecture** | New asynchronous I/O (AIO) subsystem with `io_method` setting (`worker`, `io_uring`, `sync`); up to 3× faster reads in benchmarks |
| **Vacuum** | Proactive page-freezing during regular vacuums, reducing the need for aggressive vacuum passes |
| **WAL / write throughput** | Per-connection WAL and I/O statistics for finer-grained monitoring |
| **Indexes** | Skip-scan lookups on multicolumn B-tree indexes; parallel GIN index builds; broader `OR`-clause index usage |
| **`pg_upgrade`** | Retains optimizer statistics through upgrades, reducing post-upgrade performance dips; parallel checks via `--jobs`; `--swap` flag |
| **SQL / developer features** | Virtual generated columns (default); `uuidv7()` / `uuidv4()` functions; `OLD`/`NEW` in `RETURNING` for DML statements; temporal `PRIMARY KEY`/`UNIQUE`/`FOREIGN KEY` constraints (`WITHOUT OVERLAPS`, `PERIOD`) |
| **Authentication** | OAuth 2.0 authentication; FIPS mode validation; TLS v1.3 cipher suite allowlisting; SCRAM passthrough for `postgres_fdw`/`dblink`; `md5` auth deprecated |
| **Logical replication** | Write-conflict logging; parallel streaming for subscriptions by default; `pg_createsubscriber --all` for all databases; auto-invalidation of idle replication slots |
| **Observability** | `EXPLAIN ANALYZE` auto-shows buffer access and index-lookup counts; `ANALYZE (VERBOSE)` now reports WAL, CPU, and average read stats |
| **Text processing** | `PG_UNICODE_FAST` collation; `casefold()` function; `LIKE` over nondeterministic collations; full-text search uses cluster default collation provider |
| **Wire protocol** | Protocol version 3.2 (first update since 2003); `libpq` still defaults to 3.0 for backward compatibility |
| **Data checksums** | On by default at `initdb` (off by default in 17) |


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.