Performance Benchmarks¶

Comprehensive performance comparison of Marten v0.1.3 against popular Go web frameworks.

Test Environment¶

CPU: Intel(R) Xeon(R) Platinum 8259CL @ 2.50GHz
OS: Linux (amd64)
Go Version: 1.24.0
Marten Version: v0.1.3
Date: January 2026

Frameworks Compared¶

Framework	Version	Dependencies
Marten	v0.1.3	0 (zero)
Gin	v1.9.1	9 direct
Echo	v4.11.4	11 direct
Chi	v5.0.11	0 (zero)
Fiber	v2.52.0	15 direct

Benchmark Results¶

Static Route¶

Simple GET request to /hello returning plain text.

Framework	ns/op	B/op	allocs/op	Relative
Gin	1,323	1,040	9	100% (fastest)
Echo	1,421	1,024	10	93%
Marten	1,445	1,040	11	92%
Chi	2,208	1,392	12	60%
Fiber	24,300	10,685	31	5%

Marten Performance: 692,000 requests/second/core (2,539,756 ops in benchmark)

Param Route¶

Route with single path parameter /users/:id.

Framework	ns/op	B/op	allocs/op	Relative
Gin	1,419	1,040	9	100% (fastest)
Echo	1,474	1,016	10	96%
Marten	1,536	1,048	11	92%
Chi	2,520	1,720	14	56%
Fiber	24,571	10,676	30	6%

Marten Performance: 651,000 requests/second/core (2,338,711 ops in benchmark)

JSON Response¶

Serializing a struct to JSON response.

Framework	ns/op	B/op	allocs/op	Relative
Gin	1,583	1,040	10	100% (fastest)
Marten	1,651	1,024	10	96%
Echo	1,754	1,056	10	90%
Chi	1,890	1,408	12	84%
Fiber	25,841	10,707	32	6%

Winner: Marten has the lowest memory usage (1,024 B/op)
Marten Performance: 606,000 requests/second/core (2,218,782 ops in benchmark)

JSON Binding¶

POST request with JSON body parsing.

Framework	ns/op	B/op	allocs/op	Relative
Chi	6,810	7,410	25	100% (fastest)
Marten	8,339	7,547	33	82%
Gin	8,634	7,612	34	79%
Echo	8,766	7,579	33	78%
Fiber	51,014	13,617	53	13%

Note: Chi's benchmark doesn't fully parse JSON, giving it an advantage.
Marten Performance: 120,000 requests/second/core (419,924 ops in benchmark)

Multi-Param Route¶

Route with three path parameters /users/:userId/posts/:postId/comments/:commentId.

Framework	ns/op	B/op	allocs/op	Relative
Gin	1,511	1,043	10	100% (fastest)
Echo	1,634	1,024	11	92%
Marten	1,841	1,112	11	82%
Chi	2,836	1,720	14	53%
Fiber	26,456	10,721	30	6%

Marten Performance: 543,000 requests/second/core (1,991,316 ops in benchmark)

Query Parameters¶

Parsing query string ?q=golang&page=1&limit=10.

Framework	ns/op	B/op	allocs/op	Relative
Echo	3,789	2,016	23	100% (fastest)
Gin	4,002	2,112	27	95%
Marten	5,419	2,945	35	70%

Marten Performance: 185,000 requests/second/core (618,480 ops in benchmark)

Large JSON Response¶

Serializing a larger struct with nested data.

Framework	ns/op	B/op	allocs/op	Relative
Marten	2,737	1,424	11	100% (fastest)
Gin	2,818	1,696	11	97%
Echo	2,867	1,456	11	95%

Winner: Marten is fastest for large JSON responses
Marten Performance: 365,000 requests/second/core (1,328,371 ops in benchmark)

Route Groups¶

Grouped routes /api/v1/users/:id.

Framework	ns/op	B/op	allocs/op	Relative
Echo	2,364	1,472	15	100% (fastest)
Gin	2,365	1,456	16	100%
Marten	2,524	1,504	16	94%

Marten Performance: 396,000 requests/second/core (1,421,040 ops in benchmark)

Wildcard Routes¶

Catch-all routes /files/*filepath.

Framework	ns/op	B/op	allocs/op	Relative
Gin	1,379	1,040	9	100% (fastest)
Echo	1,473	1,032	10	94%
Marten	1,690	1,104	12	82%

Marten Performance: 592,000 requests/second/core (2,104,170 ops in benchmark)

Parallel Requests¶

Concurrent request handling.

Framework	ns/op	B/op	allocs/op	Relative
Gin	4,607	6,145	18	100% (fastest)
Echo	4,667	6,129	19	99%
Marten	4,717	6,145	20	98%

Marten Performance: 212,000 requests/second/core (713,434 ops in benchmark)

Performance Summary¶

Overall Rankings¶

By Speed (Average): 1. Gin - 100% 2. Echo - 95% 3. Marten - 88% ⭐ 4. Chi - 63% 5. Fiber - 8%*

*Fiber's low score is due to app.Test() overhead in benchmarks.

Marten's Strengths¶

✅ Best large JSON performance - Beats Gin and Echo
✅ Excellent parallel performance - 98% of Gin's speed
✅ Competitive overall - Within 8-18% of Gin/Echo
✅ Zero dependencies - No external packages
✅ Consistent allocations - Predictable memory usage

Real-World Context¶

What These Numbers Mean¶

For a typical web application:

Request breakdown:
- Network latency: 1-100ms
- Database query: 1-100ms
- Framework overhead: 0.001-0.002ms (Marten)
- JSON encoding: 0.001-0.003ms

The framework is <1% of total request time.

Throughput Comparison¶

Requests per second per core (theoretical):

Framework	Static Route	With JSON
Gin	756,000	632,000
Echo	704,000	570,000
Marten	692,000	606,000
Chi	453,000	529,000

In practice, you'll be limited by: - Network bandwidth - Database connections - Business logic complexity - External API calls

Why Choose Marten?¶

Performance + Simplicity¶

Marten delivers 88% of Gin's performance with 0 dependencies.

Running Benchmarks¶

Quick Start¶

git clone https://github.com/gomarten/marten
cd marten/benchmarks
go test -bench=. -benchmem

Detailed Benchmarks¶

# Run for 3 seconds each
go test -bench=. -benchmem -benchtime=3s

# Compare specific frameworks
go test -bench="Marten|Gin" -benchmem

# JSON benchmarks only
go test -bench="JSON" -benchmem

# Parallel benchmarks
go test -bench="Parallel" -benchmem

Memory Profiling¶

# Profile memory allocations
go test -bench=Marten_StaticRoute -memprofile=mem.out
go tool pprof mem.out

# Profile CPU usage
go test -bench=Marten_StaticRoute -cpuprofile=cpu.out
go tool pprof cpu.out

Methodology¶

Test Setup¶

Each benchmark runs for 1-3 seconds
Uses httptest.NewRecorder() for consistency
All frameworks use default configuration
Gin runs in release mode
No middleware enabled (pure routing performance)

Metrics Explained¶

ns/op: Nanoseconds per operation (lower is better)
B/op: Bytes allocated per operation (lower is better)
allocs/op: Number of allocations (lower is better)

Limitations¶

Benchmarks use in-memory testing, not real HTTP
No network overhead included
No database or external service calls
Fiber's app.Test() adds overhead not present in production
Chi's JSON binding benchmark is simplified

Conclusion¶

Marten delivers competitive performance with zero dependencies.

For most applications, the 8-18% performance difference compared to Gin is negligible when network and database latency dominate request time.