The app is free · no trial Meta One · versus the free tier

Free doesn’t get less. It gets less often.

“What do you get?” assumes a feature boundary, and there isn’t one. Free and paid users run the same app, the same model, the same capabilities they differ in how many times a day they can ask the expensive questions. Which makes this a page about frequency, and a much shorter list of differences than you were probably expecting.

Free and paid tiers holding an identical set of tools, differing only in the size of the meter beneath them Free · $0 THE TOOLBOX Chat, drafting, summarising Window sharing System-wide dictation Image generation & editing Video generation HOW MANY GOES a smaller meter size not published Meta One THE TOOLBOX Chat, drafting, summarisingsame Window sharingsame System-wide dictationsame Image generation & editingsame Video generationsame HOW MANY GOES a bigger meter size not published either = The lists are identical. Only the bar underneath them changes, which is why no feature comparison can be written for this pair. Paying buys frequency, not features
Schematic the meter widths are illustrative, not measured, because neither allowance is published.
The inventory

What you get for nothing

The most useful thing this page can do is enumerate, because the free tier is broader than the word “free” suggests. Three bands, sorted by how hard each thing is metered rather than by what it does.

Free and, in ordinary use, unmetered

The things you will use every day. Nobody reports hitting a wall on these in normal desk work.

No ceiling in practice
  • The Mac app itself

    The whole 16 MB native app, not a trial or a reduced edition. Download it and sign in.

  • Chat, drafting and rewriting

    The core assistant: tightening a paragraph, drafting a reply, explaining something. Prompts worth stealing.

  • Summarising and long-document work

    Paste it or upload it and ask what changed. The context window is a property of the model, not the plan.

  • Window sharing

    Hand it one window and ask about what is on screen. How it works the permissions are the cost, not money.

  • System-wide dictation

    Types into any Mac app. The feature people keep once the novelty of the rest wears off.

  • Quick Invoke and every shortcut

    ⌥Space over any app. No shortcut is reserved for subscribers.

  • Account connections

    Instagram, Facebook and Google Workspace authorise the same way at every tier. What each grants.

  • The web version, in parallel

    Same account, same history, no separate charge. Desktop versus web.

Free, but this is where the meter lives

Available without paying and available in full the limit is on how often, and the two entries here are metered very differently from each other.

Rate-limited
  • Image generation

    Lightly metered in practice. Most people never meet the ceiling, and technique cuts attempts more than a plan does.

  • Image editing

    Each edit is a fresh generation, so a fussy round of changes costs more than one good prompt. Why that matters.

  • Video generation

    The heavy consumer, and the usual reason anyone starts paying. What video costs.

  • Long or intensive sessions

    Sustained back-and-forth can meet a limit where a handful of questions never would. What meters, in detail.

What a subscription adds on top

Short by design. One item is a bundle of things that are not Meta AI at all, one is more room, and one cannot be described.

The paid part
  • Subscriptions to other Meta apps

    The bundling is most of what Core is. Useful if you already pay for those separately the arithmetic.

  • A higher AI allowance

    More goes at the metered features above. How many more is not stated at any tier.

  • Creator and business capabilities

    Verified badges, support, scheduling, analytics, team accounts a different axis entirely. The tiers in full.

  • Premium-only capabilities

    Described as existing, not itemised. Why we will not list them.

On numbers: this page prints no allowance figures not for free, not for Core, not for Premium because Meta publishes none of them, nor the reset windows, nor what each feature costs against a limit. Sites quoting an exact daily figure are either reading something published after this page was last revised or filling the gap in. The bands above are ordered by how hard each thing meters in practice, which is observable; the sizes are not.

Three gates that don’t exist

You cannot be missing out on a capability

This is the part that makes the free tier unusually easy to recommend. The three things people assume sit behind a paywall are not behind it and each has a practical consequence.

No feature gate

There is no capability the app will show you and refuse to run because you are unpaid. Every headline feature window sharing, dictation, image and video generation works on day one at $0.

So: you can evaluate the whole product before spending anything, which is rarer than it sounds.

No quality gate

The same model answers free and paid users Muse Spark 1.3 either way. Nothing is served at reduced fidelity to unpaid accounts, and no tier produces better images.

So: a disappointing result is not a symptom of not paying, and upgrading will not fix it.

No policy gate

Content rules are policy, not a product tier. A request Meta AI declines for a free account is declined for a paying one, and money does not widen what you may generate.

So: a refusal is never a reason to subscribe it is the same answer at every price.

And nothing that was free became paid

Worth stating plainly, because the September consolidation made people nervous. When Meta One launched on 16 September 2026 it gathered existing paid tiers under one subscription; it did not move free features behind the paywall. The change that landed on the free tier was firmer rate limits on compute-heavy AI a tidy-up rather than a new bill.

Row by row

Free against paid, honestly labelled

Every row marked for what kind of difference it is. Count the “identical” labels that is the answer to the question in the title.

 Free vs Meta OneDifference
The appThe same download, the same 16 MB, the same Apple silicon requirement.Identical
The modelMuse Spark 1.3 at every tier. No premium routing, no reduced-fidelity free lane.Identical
Output qualityThe same strengths and the same failure modes. Paying buys headroom, not accuracy.Identical
Desktop featuresWindow sharing, dictation, Quick Invoke, file uploads all unpaid.Identical
Content rulesPolicy, applied the same way at $0 and $19.99.Identical
Privacy postureWhat Meta says it trains on does not change with a subscription the detail.Identical
Image and video allowanceHigher on paid tiers. This is the actual product being sold.Rate only
Session intensitySustained heavy use goes further before it meets a ceiling.Rate only
Other Meta app subscriptionsBundled by Core and above; not an AI feature at all.Added
Premium-only capabilitiesStated to exist at the top consumer tier. Not itemised publicly.Unpublished

Six rows identical, two differing only in rate, two adding something that is either not about Meta AI or not described. For the money side of this every published rate, the bundling arithmetic and what cannot be computed see Meta One pricing; for the Core-versus-Premium question specifically, the decision procedure.

The misdiagnosis

“Would paying fix this?” has two answers

Almost everyone who wonders about upgrading is reacting to a moment of friction and friction comes in two kinds that feel similar and behave completely differently.

A limit stops you. A shortfall disappoints you. Money fixes the first and does nothing at all about the second.

The reason this matters is that the second kind is far more common. Rendered text comes out garbled, a count is approximate, a real place looks nearly right but not right, the same prompt gives a different picture twice. Those are properties of the tools, present at every price. Subscribing because of them buys nothing and, worse, buries the disappointment under a sunk cost.

The test is what the screen said. If you were told you had reached a limit, that is a limit. If you got a result and did not like it, that is not and the fix is technique: edit rather than regenerate, work from a reference, or accept the thing the tool is not for. All free.

Two kinds of friction: a spent allowance, which a subscription fixes, and a disappointing result, which it does not FRICTION A · A LIMIT ! “You’ve reached your limit” the request never ran paying fixes this directly and only this Waiting also fixes it, for free limits recover. Pay when waiting is what costs you something. FRICTION B · A SHORTFALL ? the text came out garbled the request ran fine paying changes nothing same model at every tier Technique is the fix here: edit instead of regenerate, use a reference, or accept that this is not what the tool does. All of it free. Read the screen before you read the price list
Two kinds of friction, one of which a subscription touches and it is the rarer one.

Free is the right answer if…

Most people reading this page. Not a hedge the list below covers ordinary desk use almost completely.

  • You mostly use words. Drafting, tightening, summarising and asking about a window are the unmetered part of the product.
  • Images are occasional. A few a week rarely reaches a ceiling, and better prompting reaches one even less often.
  • You can wait. Limits recover. If tomorrow is fine, tomorrow is free.
  • You do not already pay Meta. Core’s bundling only saves money if you hold subscriptions to fold into it.
  • You are still evaluating. Nothing is hidden from you, so there is no evaluation you need to buy.

Still curious what you would be buying? Core versus Premium sets out the decision without guessing at the allowances.

Paying earns its keep if…

A narrower set of situations, and each one is something you can notice rather than predict.

  • You generate video regularly. The heavy workload, and the most common honest reason to subscribe.
  • You have been stopped mid-task, repeatedly. Not once, and not while experimenting during work you needed to finish.
  • An interruption costs money or a deadline. At that point $7.99 a month is priced against the delay, not the allowance.
  • You already pay for two or three Meta subscriptions. Then Core can be level or cheaper than what you hold now.
  • You need the creator or business side. Verified badges, support and analytics are a different product from allowance.

Decided to pay? Start at Core, not the top how to subscribe and what each tier costs.

FAQ

Free tier questions

What do you actually get by paying for Meta One?

More attempts at the compute-heavy features, chiefly image and video generation. Meta One is a bundle of subscriptions across Meta’s apps plus a higher AI allowance not a set of AI features withheld from free users. The full inventory.

Is the Meta AI Mac app free?

Yes, the app is free, there is no trial, and it is not a limited edition of a paid product. Chat, drafting, summarising, window sharing and system-wide dictation all work without paying anything.

What is missing from the free tier?

No headline capability is missing. What is limited is frequency: image generation, video generation and other compute-heavy work have usage limits that a subscription raises. Free users get the same tools, just fewer goes at the expensive ones.

Are the free-tier limits published?

No. Meta has not published the allowances, the reset windows or the per-feature costs at any tier, so this page does not print a number. Anyone quoting an exact daily figure is either reading something Meta published after this page was written or guessing our note on numbers.

Did anything that used to be free become paid?

No. When Meta One launched on 16 September 2026 it consolidated existing paid tiers rather than moving free features behind a paywall. The change on the free tier was firmer rate limits on compute-heavy AI, not a loss of capability the background.

Does paying make the answers better?

No. The model, the app and the output quality are identical at every tier. A subscription changes how often you can ask, not what comes back, so a wrong or approximate answer on free will be wrong and approximate on Premium too our assessment of what the tools do and do not do.

Does paying remove content restrictions?

No. Content rules are policy rather than a paid feature. If Meta AI declines a request on the free tier it will decline it on a paid tier, and treating a subscription as a way round a refusal misreads what is being sold.

How do I tell whether I need to pay?

Separate being stopped from being disappointed. If you are hitting refusals because an allowance is spent, paying fixes that directly. If the output is not good enough, paying fixes nothing that is a limitation of the tools at every tier. The two frictions, side by side.

Is the free tier enough for work?

For most desk work, yes drafting, summarising, screen questions and dictation are unmetered in ordinary use. It stops being enough when you are generating images or video in volume against a deadline, which is a narrower situation than the marketing implies.

If I do pay, which tier?

Core is the sensible first paid step at $7.99 a month, and Premium is worth considering only once Core has demonstrably stopped you. Standalone plans from $2.99 are cheaper if you want one specific thing every published rate.

Free tier · no trial, no feature gate

Start free. You are not evaluating a demo.

Six things are identical between free and paid, two differ only in how often, and nothing is hidden from you. Install it, use it for a month, and let the meter tell you whether to pay.

  • Same app, same model
  • Same quality, same rules
  • Video is what meters
  • Allowances unpublished

Independent guide not affiliated with Meta. US launch rates; tier contents and usage limits are Meta’s to state and change. Check the current terms on Meta’s own site before subscribing.