bench — HTTP Client Benchmarking Engine

Head-to-head HTTP client throughput & latency benchmarks for Bun and Node.js.

Features

• Compare 10+ HTTP clients side-by-side (RPS, latency percentiles, errors)
• Real-time terminal Grafana dashboard with sparkline charts
• Markdown table leaderboard output
• Supports Bun and Node.js
• Built-in test server on both runtimes
• Configurable concurrency, request count, and duration
• k6 scenario scripts included
• NoopGrafanaController for headless/CI runs
• Autocannon and wrk runners for cross-tool comparison

Quick Start

bun install
bun .

On Node.js:

npx tsx bench.ts

Configuration

Edit the BenchContext in bench.ts or set env vars:

Variable	Default	Description
BENCH_URL	http://127.0.0.1:3001	Target server base URL

Disabling the TUI

Switch to NoopGrafanaController in bench.ts for headless runs:

// const tui = new GrafanaController();
const tui = new NoopGrafanaController();

Enabling/Disabling Runners

Comment runners in/out of the runners array in bench.ts:

const runners = [
  bunFetchRunner,
  undiciRunner,
  hyperttpCoreRunner,
  // axiosRunner,
  // gotRunner,
  // kyRunner,
  // ...
];

Runners

All available runners. The 10 uncommented by default are marked with ✓.

Client	Source	Default
bun-fetch	src/runners/bun.ts	✓
undici	src/runners/undici.ts	✓
@hyperttp/core	src/runners/hyperttpCore.ts	✓
hyperttp	src/runners/hyperttp.ts	✓
axios	src/runners/axios.ts	✓
got	src/runners/got.ts	✓
ky	src/runners/ky.ts	✓
node-fetch	src/runners/nodeFetch.ts	✓
superagent	src/runners/superagent.ts	✓
request	src/runners/request.ts	✓
autocannon	src/runners/autocannon.ts
wrk	src/runners/wrk.ts

Metrics

Each runner runs 5 iterations; the median is reported.

Metric	Description
RPS	Requests per second
Avg	Average latency (ms)
p50/p90/p99	Latency percentiles
Min/Max	Latency range (ms)
StdDev	Latency std deviation
Errors	Failed request count

Results

Bun 1.3.14 — BunTransport

OS:  linux 7.0.10-zen1-1-zen
CPU: Intel(R) Core(TM) i5-8600K CPU @ 3.60GHz

Rank	Client	RPS	Avg	p50	p90	p99	Errors
🥇 1	bun-fetch	23.84K	8.36 ms	8.51 ms	11.01 ms	16.18 ms	0
🥈 2	undici	19.77K	10.08 ms	10.25 ms	13.79 ms	17.23 ms	0
🥉 3	node-fetch	17.85K	11.11 ms	10.09 ms	15.75 ms	25.69 ms	0
4	@hyperttp/core	11.99K	16.62 ms	15.90 ms	22.11 ms	30.79 ms	0
5	ky	11.04K	18.05 ms	18.15 ms	21.22 ms	33.33 ms	0
6	hyperttp	8.13K	24.52 ms	22.90 ms	26.67 ms	63.25 ms	0
7	request	6.92K	28.81 ms	27.79 ms	30.87 ms	44.74 ms	0
8	superagent	6.63K	30.14 ms	29.86 ms	32.41 ms	50.59 ms	0
9	got	4.41K	45.14 ms	43.63 ms	48.76 ms	74.62 ms	0
10	axios	4.37K	45.66 ms	43.70 ms	49.70 ms	66.75 ms	0

Node.js v24.16.0 — UndiciTransport

OS:  linux 7.0.10-zen1-1-zen
CPU: Intel(R) Core(TM) i5-8600K CPU @ 3.60GHz

Rank	Client	RPS	Avg	p50	p90	p99	Errors
🥇 1	undici	14.59K	13.64 ms	13.12 ms	15.13 ms	20.54 ms	0
🥈 2	@hyperttp/core	10.39K	19.12 ms	17.36 ms	22.14 ms	38.38 ms	0
🥉 3	hyperttp	9.82K	20.17 ms	18.08 ms	23.37 ms	38.70 ms	0
4	bun-fetch	7.85K	25.33 ms	23.93 ms	31.33 ms	38.77 ms	0
5	request	6.51K	30.66 ms	29.45 ms	34.83 ms	38.63 ms	0
6	ky	5.72K	34.74 ms	31.58 ms	41.56 ms	77.74 ms	0
7	axios	4.34K	45.81 ms	44.26 ms	50.27 ms	58.45 ms	0
8	node-fetch	4.20K	47.45 ms	44.75 ms	54.64 ms	71.98 ms	0
9	got	4.12K	48.49 ms	45.82 ms	55.69 ms	70.76 ms	0
10	superagent	3.00K	66.46 ms	65.06 ms	72.20 ms	83.07 ms	0

Project Structure

bench.ts          — main entry point
src/
  core/           — benchmark orchestration
  runners/        — HTTP client adapters
  tui/            — terminal Grafana dashboard
  ui/             — output formatting
  utils/          — server, TUI helpers
  types/          — shared types
  k6/             — k6 scenario scripts
  reports/        — output artifacts

License

MIT