Benchmarks

TypoKit is designed to add minimal overhead over raw HTTP servers while providing full type-safety, validation, and middleware. These benchmarks quantify that overhead across every supported platform and server adapter.

What We Measure

Scenario	Description
JSON	Serialize a fixed JSON response — measures pure routing + serialization overhead
Validate	Parse and validate a POST body — measures the validation pipeline cost
DB	Full request with SQLite read — measures realistic end-to-end latency
Middleware	Route through a 5-layer middleware chain — measures middleware dispatch overhead
Startup	Cold start to first response — measures framework initialization cost

Test Configuration

All benchmarks use bombardier with identical settings:

Connections: 64 concurrent
Duration: 10 seconds per scenario
Warmup: 3 seconds
Runs: 3 (results averaged)
Database: In-memory SQLite with identical schema and seed data

Results

JSON Serialization — Requests/sec

Logarithmic scale. Higher is better.

Validation — Requests/sec

Logarithmic scale. Includes request body parsing and validation.

Database (SQLite) — Requests/sec

Logarithmic scale. Full request lifecycle including SQLite read.

Cold Start Times (ms)

Linear scale. Lower is better. Time from process start to first response.

TypoKit Platform × Server Heatmap

JSON scenario req/s across all 13 TypoKit combinations. Rust/Axum shown in orange.

native

fastify

hono

express

axum

node

16k

17k

12k

10k

—

bun

39k

24k

15k

16k

—

deno

27k

26k

14k

19k

—

rust

—

85k

Latency Percentiles (JSON Scenario)

All values in milliseconds. Lower is better.

Rank	Framework	req/s	p50	p75	p90	p95	p99
1	raw-axum	85,321.97	1.07	1.47	1.93	2.27	3.17
2	typokit-rust-axum	84,779.28	1.09	1.48	1.94	2.27	3.14
3	competitor-elysia	53,368.87	1.87	1.97	2.21	2.43	3.37
4	raw-bun	43,222.08	2.31	2.36	2.59	2.66	2.75
5	raw-deno	42,729.57	2.33	2.39	2.46	2.50	2.61
6	typokit-node-native-cluster	41,660.73	2.21	3.08	4.17	4.94	6.78
7	typokit-bun-native	39,059.54	2.56	2.64	2.86	2.95	3.09
8	typokit-deno-native	26,649.38	3.69	3.81	3.98	4.09	4.34
9	typokit-deno-fastify	26,420.59	3.70	3.81	4.04	4.25	4.55
10	competitor-h3	25,364.17	4.03	4.18	4.29	4.34	4.67
11	raw-node	23,700.12	4.68	5.14	5.81	5.90	6.04
12	typokit-bun-fastify	23,667.87	4.57	5.04	5.30	5.57	5.98
13	typokit-deno-express	18,863.85	5.20	5.36	5.67	5.80	6.19
14	competitor-fastify	17,587.28	6.37	6.53	7.99	8.04	8.11
15	typokit-node-fastify	17,321.16	6.37	6.84	8.13	8.21	8.43

How to Read These Charts

TypoKit entries (purple) show TypoKit running on each platform × server adapter combination
Rust/Axum (orange) is visually distinguished since it’s a compiled binary, not a JS runtime
Raw baselines (gray) are zero-framework HTTP servers — the theoretical performance ceiling
Competitors (colored by framework) show popular Node.js/Bun frameworks for reference

The closer TypoKit is to the raw baseline for a given platform, the lower TypoKit’s overhead. See the Methodology page for full details on how benchmarks are run and how to reproduce them.

Running Benchmarks Locally

# Run the full benchmark suite
pnpm nx bench benchmarks

# Run only TypoKit apps
pnpm nx bench benchmarks -- --filter "typokit-*"

# Run only the JSON scenario
pnpm nx bench benchmarks -- --scenario json

# Print system info
pnpm nx bench-info benchmarks

# Print reproduce instructions
pnpm nx bench-reproduce benchmarks