📦

Drizzle ORM

orm

A TypeScript-first ORM that gives you SQL-shaped types without hiding the SQL.

orm

About

Drizzle is a TypeScript ORM designed for people who like SQL. Instead of inventing a new query language, it lets you write schema and queries in TypeScript that compile down to type-safe SQL — the generated SQL is inspectable, predictable, and as fast as what you'd write by hand. It supports PostgreSQL, MySQL, and SQLite, with first-class migrations and a Drizzle Studio GUI for browsing data. Drizzle is the opposite end of the spectrum from Prisma. Prisma generates a client from a schema language; Drizzle has no schema language — the schema is your TypeScript code. Drizzle is also noticeably faster on hot paths because it doesn't add a heavy client-side abstraction: it issues the SQL directly via the official driver. For teams that already know SQL and want their types to match it, Drizzle is the most honest ORM available in 2026.

Key Features

TypeScript schema = SQL schema
No separate DSL — your schema and your types are the same file.
Zero-overhead queries
Generates SQL directly via the official driver, no client-side abstraction.
Drizzle Kit migrations
Generate and apply SQL migrations from your schema, with no surprise queries.
Drizzle Studio
A built-in GUI for browsing and editing rows in dev.
First-class relational queries
`db.query.users.findMany({ with: { posts: true }})` with full inference.

Pros & Cons

✓ Pros

•Generated SQL is human-readable and predictable
•Performance is essentially hand-written SQL
•Schema and types live in the same file — no codegen step
•Edge runtime support is excellent (Neon, Vercel, Cloudflare)
•Migration system is honest: you see the exact SQL that will run

✗ Cons

•You need to know SQL; not a learning-curve-free ORM
•Smaller community than Prisma, fewer tutorials
•Some advanced features (e.g. groupBy with rollup) arrived later
•Type inference for deeply nested relations can be slow to compile

1 votes

Comments

emmachenJun 14, 2026

Coming from Prisma, the fact that I can read the generated SQL is a relief. No more 'why is this slow?' guessing.