Database driver popularity in 2026: a measurement across two registries

On 27 August 2026 we queried driver download counters directly from the registries. JavaScript, per week:

Python, per month:

PostgreSQL is first in both ecosystems by a wide margin — roughly one and a half to two times its nearest pursuer.

The top two positions are identical; beyond them the languages diverge:

MongoDB outranks MySQL in JavaScript and trails it in Python. That fits the history: MongoDB grew alongside Node.js and stores data in the shape the language holds it in, while the Python world carries a stronger relational tradition.

A second difference: SQL Server ranks fifth in Python and only sixth in JavaScript, at a share twenty-six times smaller relative to the leader. Enterprise integration is more often written in Python.

And a third: SQLite is absent from the Python list entirely, because there it is part of the standard library and needs no separate package. Its absence is not unpopularity but the consequence of there being nothing to download.

The most interesting movement is not between databases but within them:

Four of the five transitions are complete or nearly so. The exception is psycopg: version three shipped years ago, yet version two remains twice as popular. The reason is that it ships as a prebuilt binary requiring no compilation — a habit abandoned slowly.

Redis in JavaScript is separately curious: the unofficial client is twice as popular as the official one. That is rare, and usually means the alternative arrived first and became the habit.

Two observations. First: in JavaScript drizzle-orm has overtaken prisma — 20.0M against 16.8M. For a tool that arrived considerably later, that is a sharp reordering.

Second, and weightier: SQLAlchemy at 425.9M downloads is comparable to every PostgreSQL driver combined (530.4M). In the Python world database work overwhelmingly goes through an abstraction rather than a driver directly. JavaScript shows no such dominance: all its layers together total less than PostgreSQL alone.

Downloads are not users and not production installations. The bulk comes from build pipelines re-pulling the same thing on every run. A package fetched in every build of a microservice project will outrank one installed once on a server.

The JavaScript and Python figures cannot be compared directly — a week against a month, different ecosystem sizes and different build-caching habits.

A driver is not a database. An application may reach PostgreSQL through a layer that pulls the driver as a dependency, in which case both are counted. Conversely a database may run without a single driver download if access goes over HTTP.

Absence from the list means nothing. SQLite in Python is part of the standard library; embedded and cloud databases have no separate package at all.

What these numbers do show is relative movement within one ecosystem: driver generational shifts and changes between abstraction layers. For that they serve well, because the method’s bias applies equally to everything being compared.

To check what your database answers and how fast, use the port checker.

Sources: the npm download counter, PyPI statistics.