5 scenarios · 10 VUs · ~3 min

GraphQL Bench

NestJS — Express vs Fastify vs Mercurius

Intel(R) Core(TM) i9-9880H CPU @ 2.30GHz · 16 cores

Express

req/s on health check

avg 1.31msp99 3.39ms

Fastify

req/s on health check

avg 0.97msp99 2.6ms

Mercurius

req/s on health check

avg 0.62msp99 1.72ms

Throughput by Scenario

Requests per second — lightest to heaviest

Scenario Deep Dive

Express + Apollo

0req/s

baseline

avg 1.31msmed 1.27msp95 2.68msp99 3.39ms

Fastify + Apollo

0req/s

+32% vs slowest

avg 0.97msmed 0.93msp95 2.01msp99 2.6ms

Fastify + Mercurius

0req/s

+98% vs slowest

avg 0.62msmed 0.54msp95 1.24msp99 1.72ms

Latency Percentiles — Health Check (Scalar)

Response time distribution — lower is better

p99 Latency Across Scenarios

Worst 1% of requests — what your unhappiest users experience

Performance Radar

Normalized to fastest per scenario — 100 = best

Key Findings

Mercurius wins 5 of 5 scenarios

Up to 98% faster than slowest on light queries

Express slowest on heavy queries

23% behind fastest on Paginated Users (10)

Mutations: Mercurius wins

Express is 4% slower on writes

Gap shrinks under load

98% on Health Check (Scalar) → 23% on Paginated Users (10). Framework matters less as queries grow

Raw Data — Health Check (Scalar)

All numbers for the selected scenario

FRAMEWORK	REQ/S	VS SLOWEST	AVG	MED	P90	P95	P99	MAX
Express + Apollo	5,653	baseline	1.31ms	1.27ms	2.24ms	2.68ms	3.39ms	7.7ms
Fastify + Apollo	7,467	+32%	0.97ms	0.93ms	1.8ms	2.01ms	2.6ms	7.2ms
Fastify + Mercurius	11,191	+98%	0.62ms	0.54ms	1.16ms	1.24ms	1.72ms	4.83ms

Coming Soon

chi · Gin · Fiber

.NET + Hot Chocolate

Java

Spring + Netflix DGS