Open .tar and compressed TAR archives on Windows and extract a single file. The contents are indexed into a list first. One document is written out of a multi gigabyte server archive without unpacking the remainder.
Version 26.7 · Windows 11, Windows 10, Windows 8.1, Windows 7 and Windows Server
A TAR carries no index at the front, so every entry has to be walked before any single file can be located.
Inside the archive
What the TAR file extractor handles once the archive is loaded, before anything is written to disk.
Plain .tar sits alongside gzip, bzip2, xz and zstd. No decompressor has to be selected. The header bytes determine it before the list is built.
Free space is needed for the selected entries, not for the archive. A scheduled archive larger than the disk it is read from still opens.
Owner, group, numeric UID and GID and the permission mode are read from each header and shown per entry, which is what matters when the files are going back to a Linux host.
Symbolic links, hard links, FIFOs and character or block devices are reported as what they are rather than written out as empty files.
Entries are stored one after another, so an archive that stopped mid transfer reads cleanly up to the point it broke. Everything before the break remains selectable.
The source is opened for reading and closed unchanged. The same file passed to another party afterwards behaves identically.
The index is built once, then a 400 GB archive answers a search in the time it takes to type one.
Run the Free DemoFrom an archive too large to unpack to the single file required.
Use Open File and choose the .tar or whichever compressed form it arrived in. The archive can stay on the NAS, the external disk or the network share it arrived on. This matters where a scheduled server archive exceeds the free space on the machine reading it.
Reads across a network share, so a 400 GB archive never needs a local copy first.
Gzip, bzip2, xz and zstd each identify themselves in the first few bytes. Whether the name reads .tar.gz, .tgz, a name altered in transfer or no recognisable extension, the header determines how the file opens. No decompressor has to be selected manually.
Renamed, extensionless or double extensioned files are still identified by header.
The archive is walked from front to back and every entry recorded with its path, size, dates and Unix mode. That single pass is required by the format. Once complete, every selected file is resolved from the index rather than by a second full read.
One pass through the stream regardless of entry position.
A full server backup can hold hundreds of thousands of entries. Typing a fragment of a directory, a file name or a suffix reduces the list to a single screen, which is how one configuration file is located among the rest.
Matching runs against the whole stored path, not only the name at the end of it.
Choose a destination and only the entries you selected are written out, carrying their original sizes and modification dates. A three megabyte configuration file consumes three megabytes of disk. The four hundred gigabyte archive containing it is left untouched.
The source file is only ever read, never altered.
Paths from a Linux server run past the 260 character MAX_PATH limit Windows applies to ordinary extraction. A rejected write at the end of a long job is the usual way this surfaces. Exporting to something like D:\out keeps you clear of the limit whatever the original tree looked like.
The full original path stays visible in the list either way.
More than the files. It carries the Unix world they came from, which is where Windows starts losing things.
Every entry records its Unix mode, its numeric owner and group and its modification time. Windows has no accounts to map those onto, so they survive as information rather than as access rights.
Symbolic links, hard links, devices and named pipes all have their own record types. They are listed as what they are instead of quietly turning into empty files.
Linux is case sensitive and Windows is not, so Notes.txt and notes.txt can share one archive. Both appear in the list, which is the only way to choose the one you meant.
Yes, on Windows 11. That runs out the moment a TAR is a backup rather than a download.
File Explorer has read .tar and its compressed variants since Windows 11 version 23H2, through the libarchive project. Version 24H2 made it the documented default. Double click a 40 MB source archive and you are done. Nothing here argues with that.
The gap is not the format. It is the size, the state and the machine. Explorer does none of it on Windows 10. A TAR keeps no index either, so a 400 GB nightly cannot be opened the way a 40 MB download can.
Three of these cost nothing and two ship with Windows. Each has a different limit.
| Route | On Windows 10 |
One file at a time |
Reads a cut short file |
Shows Unix metadata |
|---|---|---|---|---|
| File Explorer Free Built into Windows 11 from 23H2 |
||||
| The tar command Free The bsdtar build, shipped with Windows since 2018. In Command Prompt or PowerShell, tar -xvzf archive.tar.gz -C D:\out |
||||
| An archive tool or WSL Free Right-click extraction or the same tar command inside Ubuntu on WSL |
||||
| TAR file extractor Demo Indexes once, exports what you tick |
On Windows 11 with a small archive, File Explorer is the right answer and costs nothing. The tar command handles more but expects the syntax and gives you no list to search. Right-click extraction is built for downloads rather than for a 400 GB nightly. Web based converters get recommended a lot and are worth avoiding here, because uploading a server backup puts every file in it on hardware you do not control. The TAR extractor tool is for the archives the rest find awkward.
TAR is rarely selected deliberately. It arrives as the output of another process, typically on an unattended schedule.
NAS appliances write them. Hosting panels package an account into one on demand. Backup schedulers and rsync scripts produce a nightly and drop it on a share. Container images and appliance exports are tar underneath. A Linux server being decommissioned usually leaves one final archive of everything that mattered.
In each case the machine that wrote the archive is typically no longer available, and the archive is being opened on Windows. The archive itself is intact. What is missing is the original environment, along with the disk space to unpack a full copy in order to read one configuration file.
TAR sits in the entry tier, so the cheaper edition is the right one here.
Professional, $99
Relevant only where Veeam, Acronis, ShadowProtect, tape or virtual disk images are also held. TAR alone does not require it.
Run the TAR extractor software against your own archive first. It builds the index and displays the required entry before any purchase.
Buy Standard Compare all editions
Accurate for Backup Extractor 26.7. BitRecover has been building data recovery and format conversion tools since 2011.
Most people meet this the hard way. It is a property of the format rather than a limitation of any tool.
A TAR has no central directory. A tarball is a plain sequence of records, each one a 512 byte POSIX ustar header followed by the file it describes, laid end to end in the order they were written. Add a compressor around that and the cost of reaching anything changes again.
One passNo structure records entry positions, so the archive is read through once to establish them. Each position is stored during that pass and every later selection resolves directly to its offset.
Every byteGzip, bzip2 and xz wrap the whole container in a single stream. There is no position to jump to, so reaching that file means decompressing all 292 gigabytes ahead of it. A data error part way through stops everything after it.
The same layout carries one advantage. Because TAR is a sequence rather than a structure, an archive that was cut short is still readable up to the point it stopped. A download that failed at 80 percent or a tape that ran out mid write is not a total loss. The file list will show you exactly how far it got.
Common questions about large archives and limited destination space.
Windows 11 can. File Explorer reads .tar and the compressed variants from version 23H2 onward. Windows 10 cannot do it from Explorer, only from the tar command in a terminal.
Because gzip compresses the whole archive as one continuous stream. Reaching a file 300 GB in means decompressing the 300 GB in front of it first. A plain .tar has no such problem.
Yes. The TAR file extractor indexes the archive first, then writes out only the entries you tick. Nothing else lands on your disk.
The gzip, bzip2, xz and zstd variants, so .tar.gz, .tgz, .tar.bz2, .tar.xz and .tar.zst all load the same way as a plain .tar.
Often yes. A TAR is a sequence of file records, so a partial archive still reads up to the damaged point even when the tail is gone. Load it in the TAR file extractor and the list shows how far it got.
They are stored in the archive and shown against each entry. Windows has no matching user accounts, so the files are written out with Windows permissions and the original values are there to read.
Linux treats Notes.txt and notes.txt as two files and Windows does not. Both are listed separately so you can pick the one you want rather than have the second overwrite the first.
Paths from a Linux server often run past the limit Windows applies to ordinary extraction. Choosing a short destination folder is the simplest fix. The file list shows the full original path either way.
No. Reading and export are one directional. The .tar or .tar.gz on disk is the same file afterwards, byte for byte.
Standard at $49. TAR is an entry tier format. Professional adds nothing required for TAR archives.
Load the archive into the TAR file extractor, search the list and export only the required entry.
Version 26.7 · Windows 11, Windows 10, Windows 8.1, Windows 7 and Windows Server · 30 day money back