Edit distance: .NET ArrayPool is 71× faster than Python textdistancePageRank on 28.5M edges: NetworkX takes 47× longer than .NET CSRInterval tree: augmented BST answers 10k queries 31× faster than PythonFull-text search: Lucene.NET handles 22× more queries/sec than WhooshDiff algorithm: preallocated DP LCS is 21× faster than difflib at 50k pairsDate parsing: DateTimeOffset is 19× faster than dateutil at 1M timestampsQR codes: QRCoder generates 50k codes 14× faster than Python qrcodeMarkdown: Markdig renders 10k documents 11× faster than mistuneTokenizer: compiled regex tokenizes 100 MB 8× faster than NLTKPDF extract: PdfPig extracts 200 papers 6× faster than pypdfSoftmax: TensorPrimitives is 1.53× faster than NumPy at Llama-3 vocab scaleEdit distance: .NET ArrayPool is 71× faster than Python textdistancePageRank on 28.5M edges: NetworkX takes 47× longer than .NET CSRInterval tree: augmented BST answers 10k queries 31× faster than PythonFull-text search: Lucene.NET handles 22× more queries/sec than WhooshDiff algorithm: preallocated DP LCS is 21× faster than difflib at 50k pairsDate parsing: DateTimeOffset is 19× faster than dateutil at 1M timestampsQR codes: QRCoder generates 50k codes 14× faster than Python qrcodeMarkdown: Markdig renders 10k documents 11× faster than mistuneTokenizer: compiled regex tokenizes 100 MB 8× faster than NLTKPDF extract: PdfPig extracts 200 papers 6× faster than pypdfSoftmax: TensorPrimitives is 1.53× faster than NumPy at Llama-3 vocab scale

.NET

.NET is seriously fast!

We rewrite the world's most popular code and show the benchmarks. Real numbers. Real wins. Sub-millisecond proof.

See top benchmarks Compare code

RECORD OF THE TIME

May 2026

CSV Processing 100M rows

stream-parse · cold start · single thread

.NET 94.1 s

Node 2289.0 s

Ruby 3.3336.0 s

82x

faster

.NET MarkdigPy mistune

.NET.NET

800 ms

PyPython

8.8 s

11×

vs PYTHON

See all 11 benchmarks

SIDE BY SIDE

Same job. Different worlds.

Parse a 10k-object JSON payload. Watch the gap.

.NET.NET 9 / C#

1.2 ms

1// .NET 9 — System.Text.Json (source-gen)
2using System.Text.Json;
3
4[JsonSerializable(typeof(Order[]))]
5partial class Ctx : JsonSerializerContext { }
6
7var sw = Stopwatch.StartNew();
8var orders = JsonSerializer.Deserialize(
9    json, Ctx.Default.OrderArray);
10sw.Stop();
11// 1.2 ms  •  0 allocations on hot path

PyPython 3.13

56.4 ms

1# Python 3.13 — json stdlib
2import json, time
3
4t = time.perf_counter()
5orders = json.loads(raw)
6elapsed = (time.perf_counter() - t) * 1000
7
8# 56.4 ms  •  GC pressure
9# ~47x slower than .NET 9
10# (and that's before model binding)

COMMUNITY

Developers are talking.

Submit a benchmark. Pick apart ours. Move the numbers.

Alex Kovalev

Senior backend, ex-Google

I migrated a Python data pipeline to .NET 9 and it shaved 38 minutes off our nightly job. Source-gen JSON is unreal.

Priya Raman

Indie game dev

AOT + NativeAOT means I ship a single 6MB binary that boots in 4ms. This site keeps proving what I tell my team daily.

Marco Bianchi

Platform engineer

Was a Go diehard. The regex benchmark made me actually try it. Now half our edge services are .NET minimal APIs.

Sara Lin

OSS maintainer

Submitted a Rust vs .NET image convolution PR — got merged in two days. Best community I've contributed to in years.

Contribute on GitHub