Prisma vs Supabase vs Drizzle Benchmark

Query performance comparison using a real-world multi-join query: conversations + contact_channels + contacts + workflows + checkouts + workflow_goals + messages.

Results

1000 iterations, concurrency 10

	Avg	P50	P95	P99	Min	Max
Raw SQL	4.4 ms	3.8 ms	9.8 ms	17.7 ms	0.8 ms	22.5 ms
Drizzle	11.1 ms	10.9 ms	17.0 ms	28.6 ms	3.0 ms	64.4 ms
Prisma	35.2 ms	34.8 ms	46.7 ms	58.5 ms	20.0 ms	92.7 ms
Supabase	55.5 ms	56.1 ms	69.5 ms	103.2 ms	27.6 ms	128.4 ms

Winner: Raw SQL (51.1 ms / 92.1% faster than Supabase)

How it works

File	Role
server.ts	Express server exposing GET /prisma, GET /supabase, and GET /drizzle endpoints
bench.ts	Fires 1000 requests at each endpoint (10 concurrent), collects server-side timings, reports percentiles

Both endpoints query the same PostgreSQL database, returning the latest 100 conversations with all related data.

Setup

# install
npm install
npx prisma generate --schema=./prisma

# configure .env
DATABASE_URL=postgresql://user:pass@localhost:5432/dbname
DATABASE_API_URL=http://localhost:8080
JWT_SECRET=your-jwt-secret

# run
npm run server   # start the benchmark server
npm run bench    # run the benchmark (server must be running)

Stack

• Prisma - TypeScript ORM with findMany + include for joins
• Drizzle - Lightweight TypeScript ORM with SQL-like query builder
• Supabase PostgREST - REST API over PostgreSQL with embedded select syntax
• Raw SQL - Direct queries via postgres.js (tagged template literals)
• Express - HTTP server
• tsx - TypeScript execution