What is in the release
The Document Foundation released LibreOffice 26.8 on 26 August 2026. The release notes lead on three things, none of which are AI.
Typography. A new Paragraph Composer balances word spacing across consecutive lines rather than optimising each line independently. That is the difference between text that looks acceptable line by line and text that looks even as a block, and it is the kind of thing that separates a document that reads like a document from one that reads like output.
Writing systems. Automatic paragraph direction detection, improved right-to-left handling, and support for the N'Ko and Adlam scripts in Math.
Document exchange. Round-trip support for newer chart types Microsoft Office writes into OOXML files, including waterfall and treemap charts. Round-trip is the operative word: opening someone's file and returning it without quietly destroying the charts.
The part they wrote down deliberately
The release notes state plainly: "LibreOffice 26.8 contains no generative AI features. Documents are not transmitted to remote services for processing, and no component of the suite requires network access to function."
That is a positioning statement, and an unusual one in a year when every commercial office suite has been adding assistants and charging for them. Stating the absence as a property rather than omitting to mention it is a decision about who the software is for.
For most buyers it is irrelevant or a drawback. For a specific and not small set of them, legal practices, healthcare, defence work, anyone under a data residency obligation, anyone whose documents are covered by a confidentiality regime that does not care how good the vendor's data handling policy is, a suite that cannot transmit a document is easier to approve than one that promises not to.
What this does not settle
Two things worth being clear about, because a release note is not an evaluation.
The absence of AI is not by itself an argument for switching. Microsoft 365 and Google Workspace are ahead on collaboration, real-time co-editing and mobile, and those gaps are the reason most organisations that considered LibreOffice stayed where they were. A no-AI position does not close them.
And OOXML round-trip support improving is not the same as OOXML round-trip support being complete. Complex documents moving between suites still lose formatting, and any organisation exchanging heavily formatted files with Microsoft Office users should test with their own documents rather than take a release note's word for it.
Who a suite that cannot phone home is actually for
The instinct is to read a no-AI release as a project falling behind. For a large share of buyers that reading is correct and they should use something else. For a specific set it is backwards, and the distinction is about procurement rather than preference.
Consider what approving a cloud office suite involves in a regulated environment. Someone has to establish where documents are processed, which sub-processors touch them, what the retention is, whether training on customer content is possible and under which contract terms it is excluded, and what happens to all of that when the vendor updates its terms. That review is expensive, it has to be redone when the vendor changes anything, and it ends in a judgement about a policy rather than a fact about the software.
A suite that requires no network access to function ends the review differently. The question stops being what the vendor does with the document and becomes whether the document can leave at all. That is a property that can be verified rather than a commitment that must be trusted, and verified properties are dramatically cheaper to approve.
This is why LibreOffice persists in defence, government, healthcare and legal work well past the point its feature set would otherwise justify. It is not usually chosen because it is better. It is chosen because it is approvable.
What the licence saving is, and what it is not
LibreOffice has no per-seat licence, and for organisations at scale that number is large enough to be worth stating plainly rather than gesturing at.
What it does not mean is that switching is free. The costs simply move, and they move to places that are harder to see in a budget line. Retraining is real, particularly for people whose work lives in advanced spreadsheet features, where the gap between LibreOffice Calc and Excel is widest and where the affected users are often the least willing to change. Macro and template migration is real: an organisation with years of accumulated VBA does not port it in an afternoon, and some of it will not port at all. Support becomes an internal responsibility or a contract with a third party rather than something included.
The organisations that have made this switch successfully share a pattern. They moved departments where document work is straightforward, kept licences for the finance and analyst teams who genuinely need Excel, and treated it as a multi-year programme rather than a cutover. The ones that failed tried to move everyone at once and discovered the spreadsheet problem in month two.
So the saving is real and it is not the whole calculation. Anyone presenting a per-seat multiplication as the business case is describing a best case that has rarely happened.
Is shipping without AI a strategy or a shortfall?
Both readings are available and it is worth separating them, because the answer matters for anyone betting on this suite for the next five years.
The strategic reading: The Document Foundation is a non-profit funded by donations. Building competitive generative AI features means either running inference, which costs money it does not have, or sending documents to a third party, which contradicts the reason a large part of its user base chose it. Declining to do either is a coherent decision that plays to a real and defensible position rather than chasing a fight it would lose.
The shortfall reading: they could not do it, and the release notes make a virtue of the constraint. Foundations move slowly, the resourcing gap against Microsoft is enormous, and stating the absence prominently is what you do when the absence will be noticed anyway.
The evidence mildly favours the first. Writing it into the release notes as an explicit property, alongside the claim that no component requires network access, is a positioning statement rather than an omission, and it is aimed at an audience that exists and buys on exactly that basis. But the second reading is not unreasonable, and anyone choosing this suite should understand that the AI gap will widen rather than close, because the economics that produced it are not changing.
What would settle it is what happens if a local-inference option becomes cheap enough to ship without network access. That would satisfy both the positioning and the capability gap, and whether the project pursues it will say more about the intent than this release does.
The honest comparison with Microsoft 365 and Google Workspace
A directory that only listed the case for LibreOffice would be useless, so here is the case against, which is substantial.
Collaboration. Real-time co-editing in Microsoft 365 and Google Workspace is mature and genuinely good. LibreOffice has collaborative editing through third-party integrations and it is not comparable. For an organisation whose documents are worked on by several people at once, this single gap usually decides the question.
Mobile. Microsoft and Google both ship capable mobile apps. LibreOffice's mobile story is weak, and for a workforce that reviews documents on phones that is disqualifying.
Fidelity. Complex Microsoft Office documents still lose formatting in transit. 26.8 improves chart round-tripping specifically, which is real progress on a real problem, but anyone exchanging heavily formatted files daily will still hit differences.
Administration. Central policy management, conditional access and audit tooling are areas where the commercial suites have invested heavily and LibreOffice has not.
The fair summary is that LibreOffice wins on cost, on document ownership and on approvability, and loses on collaboration, mobile and administration. Which set matters is not a question a review can answer for you, and any comparison that declares an overall winner is not describing your situation.
The quieter argument: who owns the file in ten years
Underneath the AI question sits an older one that does not get raised often and matters more for some organisations than any feature comparison.
LibreOffice's native format is OpenDocument, an ISO standard implemented by more than one vendor. Microsoft's OOXML is also a standard, and is in practice implemented completely by one product. The distinction is thin in normal operation and becomes load-bearing in exactly the situations where it is too late to change: a records retention obligation measured in decades, an archive that must remain readable after a vendor relationship ends, a legal discovery process reaching back years.
Public sector bodies and archives have been making this argument for twenty years, which is why several European governments have standardised on OpenDocument independent of any view about cost or AI. The argument is not that Microsoft will stop reading its own format. It is that a format whose complete definition lives with one vendor carries a dependency that outlives any particular contract, and some organisations are required to care about that.
For most businesses this is not a consideration and pretending otherwise would be dishonest. For anyone with a statutory retention duty, it is arguably the strongest argument in favour, and it has nothing to do with what shipped in 26.8.
If you are evaluating it, test these four things
A release note cannot tell you whether a suite works for your organisation. Four tests will, and all of them use your own documents rather than a demo.
Your worst spreadsheet. Not a representative one, the one with the pivot tables, the array formulas and the macro somebody wrote in 2019 who has since left. If it survives, the rest will. If it does not, you have found the department that keeps its Excel licences, and that is a useful answer rather than a failed test.
A document that goes out to clients. Open it, edit it, save it back to the Microsoft format, and open it again in Microsoft Office. Round-tripping is where fidelity problems appear, and 26.8 specifically improves chart round-tripping, so a document with waterfall or treemap charts is the sharpest version of this test.
Your template set. Branded templates carry styles, headers and often macros. This is where migration effort concentrates and where estimates are usually wrong by the largest margin.
A genuinely collaborative document. Have two people try to work on the same file at the same time, the way they actually do. This is the test LibreOffice is most likely to fail, and failing it early is worth more than discovering it after committing.
Run those four before reading another comparison. They will tell you more about your own situation than any review can, including this one.
Why the typography change is more interesting than it sounds
The Paragraph Composer is the kind of feature that gets skipped in coverage and is disproportionately visible in output.
Most word processors set each line independently: fit the words, adjust the spacing, move on. The result is that one line can be tightly spaced and the next loose, and while no single line looks wrong, the block has an uneven texture people notice without being able to name. Professional typesetting has balanced spacing across consecutive lines for decades, which is part of why a typeset page reads differently from a word-processed one.
Bringing that into a free office suite closes a gap with professional layout tools for anyone producing documents meant to be read at length rather than circulated and archived. Combined with the writing systems work, automatic paragraph direction detection and support for the N'Ko and Adlam scripts, the release is aimed at people producing finished documents rather than at people collaborating on drafts.
That is a coherent position, and it matches the no-AI decision rather than sitting oddly beside it. Both are choices about what the software is for.
Questions people ask
- Does LibreOffice 26.8 have any AI features?
- No. The release notes state it contains no generative AI features, that documents are not transmitted to remote services for processing, and that no component requires network access to function.
- Is LibreOffice free for commercial use?
- Yes. LibreOffice is free and open source, including for commercial use, with no per-seat licence. The Document Foundation is funded by donations and sponsors.
- Can it open and return Microsoft Office files without breaking them?
- 26.8 adds round-trip support for newer OOXML chart types including waterfall and treemap. Round-trip on complex documents is still imperfect between any two suites, so test with your own files before committing to an exchange-heavy workflow.
- Who is the no-AI, no-network position actually for?
- Organisations where documents cannot leave the premises, legal, healthcare, defence, and anyone under data residency rules. A suite that cannot transmit a document is easier to get approved than one that promises not to.
Sources
Every price, limit and date above was checked against these pages on the day shown. Where a figure has since moved, the vendor’s own page is the authority and this one is a snapshot.
- 1LibreOffice 26.8 release announcement · The Document FoundationPrimary. Released 26 August 2026. States the suite contains no generative AI features, that documents are not transmitted to remote services for processing, and that no component requires network access to function. Lists the Paragraph Composer, writing systems work and OOXML chart round-trip support.Checked