PostgreSQL
PostgreSQL is stewarded by the PostgreSQL Global Development Group (PGDG), a volunteer, vendor-neutral collective with no single corporate owner — a structural difference from Oracle-owned MySQL. Current stable release is PostgreSQL 18 (shipped September 25, 2025, patched to 18.4 in May 2026); PostgreSQL 19 hit beta 2 in July 2026. It ships under the OSI-approved PostgreSQL License, a permissive MIT/BSD-style grant with no copyleft. Technically, it combines MVCC concurrency with a new v18 asynchronous I/O subsystem — defaulting to a portable worker-process method, with an optional Linux-only io_uring backend — delivering roughly 2-3x faster sequential and bitmap scans on I/O-bound workloads (biggest gains on network-attached cloud storage, more modest on fast local SSD). Adoption: the 2025 Stack Overflow survey found 55.6% of all developers and 58.2% of professionals use it — the most-used database, and most admired/desired since 2023 — even as DB-Engines still ranks it #4 by raw score behind Oracle, MySQL, and SQL Server. Current version/status: 18.4 (major version 18 released 2025-09-25; patch 18.4 released 2026-05-14; PostgreSQL 19 in beta as of July 2026). License: PostgreSQL License — OSI-approved, permissive MIT/BSD-style, no copyleft. Pricing: Core database is free, open-source software with no license fees for self-hosting. Managed hosting pricing varies widely: Supabas
Pick PostgreSQL when the app needs relational integrity plus workloads that don't fit clean rows — JSONB documents, geospatial via PostGIS, full-text search, or vector embeddings via pgvector — without standing up three separate databases. Skip it for simple key-value caching or massive-scale time-series ingestion, where Redis or a purpose-built store like TimescaleDB/ClickHouse will outrun a general-purpose RDBMS. The community's own honest gripe: native horizontal write-sharding is still immature, so scaling writes across nodes means bolting on Citus or hand-rolling partitioning — Postgres wins on breadth, not turnkey scale-out. Known weakness: No built-in horizontal write-sharding — scaling writes across nodes requires third-party extensions like Citus or manual partitioning, a gap the project itself acknowledges versus natively distributed databases
See how it ranks →