Skip to content
Zahin Afsar
2 min read

Best Way to Query Postgres in Node.js — Prisma vs Drizzle vs Supabase vs Raw SQL

Four ways to query Postgres from Node. Pros, cons, and benchmark numbers from a real multi-join query.

By Md. Zahin Afsar
#postgres#nodejs#prisma#drizzle#supabase#sql#benchmark

Four real options: Prisma, Drizzle, Supabase PostgREST, raw SQL (postgres.js). I benchmarked all four with the same multi-join query. Here's what wins, what loses, and when.

Repo: github.com/zahinafsar/node-orm-benchmark

Benchmark

Same Postgres. Same query. 100 rows joined across 7 tables (conversations + contacts + messages + workflows + checkouts). 1000 iterations, concurrency 10.

Latency by method — Avg, P95, P99. Lower is better.

Raw SQL is 2.5× faster than Drizzle, 8× faster than Prisma, 12× faster than Supabase. Tail latency (P99) is where the gap really widens.

1. Raw SQL — postgres.js

ts
const rows = await sql`
  SELECT c.*, cc.username, co.given_name
  FROM conversations c
  INNER JOIN contact_channels cc ON cc.id = c.channel_id
  INNER JOIN contacts co ON co.id = cc.contact_id
  LIMIT 100
`;

Pros

  • Fastest. Nothing between your query and Postgres.
  • Full SQL — window functions, CTEs, anything.
  • No abstraction to learn.

Cons

  • No type safety on results.
  • No migrations.
  • Typos fail at runtime, not compile time.

Use when: hot endpoints, reports, anything perf-critical.

2. Drizzle

ts
await db
  .select()
  .from(conversations)
  .innerJoin(contact_channels, eq(conversations.channel_id, contact_channels.id))
  .limit(100);

Pros

  • Almost as fast as raw (2.5× slower).
  • Typed queries, typed results.
  • One SQL statement per query — no N+1 surprises.
  • Lets you drop to raw SQL anytime.

Cons

  • API mirrors SQL — if you don't know SQL, the learning curve is real.
  • Younger ecosystem than Prisma.

Use when: default pick for new projects.

3. Prisma

ts
await prisma.conversations.findMany({
  take: 100,
  include: { contact_channels: { include: { contacts: true } } },
});

Pros

  • Best DX. Schema-first, autocomplete everywhere.
  • Mature migrations, Studio GUI.
  • Huge ecosystem.

Cons

  • 8× slower than raw on multi-joins.
  • include issues multiple queries and stitches in Node — not one SQL statement.
  • Separate Rust/WASM engine adds FFI overhead.

Use when: DX matters more than perf, mostly simple queries.

4. Supabase PostgREST

ts
await postgrest.from('conversations').select('*, contact_channels(*)').limit(100);

Pros

  • Auto-generated REST API.
  • Row-level security baked in.
  • Great for browser → DB directly.

Cons

  • Slowest. Extra HTTP hop: Node → PostgREST → Postgres.
  • JWT verification per request.
  • Wrong tool when you already have a Node server.

Use when: browser client talking to Postgres. Never from a Node backend.

TL;DR

  • Default: Drizzle.
  • Hot paths: raw SQL via postgres.js. Drizzle lets you mix both.
  • Avoid: Prisma if P99 latency matters. PostgREST if you're already in Node.

Try it yourself:

bash
git clone https://github.com/zahinafsar/node-orm-benchmark
npm install && npm run server
npm run bench