DuoBolt vs Czkawka — Speed, Safety and Restoring Files

Czkawka is a popular open-source duplicate finder, and on speed it is the closest competitor DuoBolt has. How far apart they land depends on what is limiting the scan: wide where per-file work decides it, a few percent on an external disk, and nothing measurable over a network.

Benchmark Results

All tests ran on an Apple M2 Max, macOS 27.0, on 20 September 2026, with each tool’s own cache cleared before every run, a 1 KB minimum file size and exact duplicates only. Czkawka 12.0.2, the czkawka_gui GTK build — the one brew install czkawka produces. Lower times are better.

Storage	DuoBolt	Czkawka	DuoBolt advantage
Internal SSD — 394 GiB, 1.69M files	38s	1m3s	1.66× faster
USB-C SSD — 797 GB	5m17s	5m39s	1.07× faster
NAS over SMB — 1.09 TiB	10s	11s	Tie

Key takeaway: the gap tracks what is limiting the scan rather than how large the dataset is. Across 1.69 million small files, where the cost is per-file work, DuoBolt finishes in well under half the time. On an external SSD the bus sets a ceiling both tools run into and the difference falls to 7%. Over SMB the two land inside the measurement’s resolution of each other. Of the four tools measured against DuoBolt, Czkawka is the closest on every tier.

The project now directs new users to Krokiet, a separate Slint front end sharing the same czkawka_core. The GTK build was measured because it is what Homebrew installs today. The full method covers all five tools.

Feature Comparison

Feature	DuoBolt	Czkawka
Platforms	macOS, Windows	macOS, Windows, Linux
Price	Free tier · $34.99 one-time	Free (open source)
Sold as a subscription	No	No
Hashing	BLAKE3	BLAKE3 by default, or CRC32 / XXH3
Similar photos	Yes, with AI Curation	Yes
Adjustable similarity	Yes — five bands, Identical to Loosely similar	Yes — similarity presets
Opens RAW files	Yes — CR2/CR3, NEF, ARW, RAF, ORF, RW2, DNG and more	Depends on the build’s codecs
HEIC / AVIF	Yes (Windows needs two free Store extensions for previews)	Depends on the build’s codecs
APFS clone awareness	Yes	No
NAS / network drives	Yes	Yes
Rescan cache	Yes — hashes and thumbnails, separate local and network thresholds	Yes
Re-runnable scan history	Yes — each scan’s full setup is stored and can run again	—
Safe deletion	Trash, archive, or copy	Delete, or Trash with “Move deleted files to trash” enabled
Review before removing	Mandatory Review Selection step, counting what would be left with no copy	Not in its documentation
Protected folders	Yes — files stay unselectable	Excluded directories, applied at scan time
Recovers permanent deletions	Yes — rebuilt from a byte-identical copy, with the original dates and tags	—
Account / telemetry	None	None

When to Choose DuoBolt

You want a review step before removing

DuoBolt gives you a visual interface to review and preview before anything moves, five ways out — Trash, delete, archive, copy and, for duplicates on a Mac, replace with a clone — and a history that restores from a surviving copy afterwards.

You scan millions of small files

Where per-file work decides the result rather than raw throughput, DuoBolt finishes first: 38s against 1m3s across a source tree of 1.69 million files.

You want a removal to be undoable

DuoBolt never removes anything without a confirmation screen, and Operations History can restore what it moved from a surviving copy.

When to Choose Czkawka

You need a free, open-source tool
You work primarily on Linux
You prefer a tool with a larger community and public issue tracker
Your scans are small, where the time difference is a few seconds

Weighing Czkawka against the other free option instead? Czkawka vs dupeGuru has both measured side by side.

Where DuoBolt’s Time Goes

Both tools hash with BLAKE3 by default, and both prehash before committing to a full read, so the gap in the table above is not down to the algorithm. What DuoBolt does with it:

Per-root parallelism — all CPU cores stay busy across multiple scan roots at once, rather than one root after another
Streaming I/O — chunked reads overlap disk access with hashing, so a core is not waiting on a slow or network drive
Two-stage prehash — first and last 128 KiB, dropped to a 4 KiB head-only read on network volumes, where every extra round trip costs more than the bytes do

We have not read Czkawka’s source to say which of these it does differently. The timings are ours and the methodology is published; the explanation above is about our engine, not a claim about theirs.

Full Benchmark Results

All three storage tiers, compared against Czkawka, dupeGuru, Gemini 2 and Duplicate File Finder.

View Benchmarks →

Exact duplicate file finder

Why the duplicate engine only flags byte-for-byte matches — no name shortcuts, no fuzzy logic, and what that verdict does and does not claim.

Download DuoBolt

Free to scan on macOS and Windows, with a free tier that also cleans.

Download →