# Meta monthly active users: Community Rights End at 700,000,001

Priya Menon · October 1, 2026

> Meta's 700,000,001 MAU cutoff is contractual: Community terms end and a Llama download receipt becomes a separate commercial license, not a capability test.

| Takeaway | Detail |
| --- | --- |
| The trigger is contractual. | Above the applicable MAU line, the published Community terms give way to a separate Meta commercial grant. |
| The required artifact changes at the cutoff. | Above the applicable MAU line, a download receipt for the referenced Llama model file is replaced by a separate commercial license signed by Meta. |
| The cutoff is not a capability benchmark. | The large-user MAU threshold identifies a contractual escalation point, not a model-size test or public price tier. |
| The public acquisition path is unspecified. | Above the applicable MAU line, the separate Meta commercial grant has no standard public application form or fee disclosed. |

The surprising issue is the monthly active-user threshold. In the licensing framework described here, that threshold is not a model-size milestone or a familiar API price tier. It is the point at which the published Community terms cease to provide the relevant license and a separate Meta commercial grant is required.

That distinction changes the artifact an eligible user must keep. Below the stated threshold, the account can rely on the Community license and a download receipt for the referenced Llama model file. Above the applicable MAU line, the hook instead points to a separate commercial license signed by Meta. The receipt may document acquisition, but it does not substitute for the grant.

The practical question is therefore not whether a product has passed some technical capacity benchmark. It is whether the account has crossed the contractual line and obtained the required document. The supplied material identifies no standard public application form or fee for that commercial grant, so teams should not treat an ordinary download confirmation as evidence of commercial authorization.

![Meta monthly active users](https://static.mm-ais.com/article-images-ai/meta-monthly-active-users-community-righ-ai-283b8595.jpg)

## Large-User MAU: Where Community Rights End

Meta Platforms, Inc. draws a contractual—not operational—line under its public Llama 3.1 Community License Agreement. For 2026 deployments, Meta is the licensor, and that agreement is the governing text for the Llama 3.1 weights used in the commercial product. According to Meta’s clause, the trigger is a monthly active-user count above the license-defined line. Applied to the commercial product or service using the model, it is a strict inequality: the line itself is outside the clause, while the next monthly active user is inside it.

| Current license-relevant MAU | Result under Meta’s clause | Approval control |
| --- | --- | --- |
| At the line | Does not trigger the large-user clause | No escalation under this clause; all other license requirements still apply |
| Above the line | Triggers the large-user clause | Do not approve launch, expansion, or continued operation at that scale until counsel holds a signed separate Meta commercial license covering the use |

The escalation instrument has a different legal character: **a commercial license from Meta**. It is a separate written grant covering the above-line commercial use. A second download, duplicate copy, or renewed click acceptance of the Community License is not that grant. The license register must therefore treat the Community License and the separate Meta grant as distinct legal instruments, not interchangeable evidence of authorization.

Preserve the unit of account. The comparison metric is the current monthly active-user count of the commercial product or service using the model. Model-parameter counts describe architecture; tokens processed, API calls, conversations, and concurrent sessions describe workload, traffic, or capacity. None substitutes for monthly active users in this clause. A dashboard may collect those operating metrics alongside the user count, but counsel’s threshold decision must remain attached to the product or service’s MAU field.

For each product register, record whether the deployment uses Meta’s released weights, a fine-tuned Derivative Model, or an embedded or API feature. Carry that designation, the governing agreement, and the commercial-license escalation status into the product’s license register. The control is linkage: before counsel clears a launch or expansion gate, the signed Meta grant must be matched to the deployment scope actually recorded. A generic entry saying only “Llama” is too imprecise to support that review.

![Large-User MAU: Where Community Rights End — Meta monthly active users](https://static.mm-ais.com/article-images-ai/meta-monthly-active-users-community-righ-ai-44826399.jpg)

## 8B and 70B

Parameter count is a versioning fact, not a license metric. Model size does not establish a product’s production MAU, identify the license governing a derivative, or authorize operation above Meta’s contractual line.

| First-party record | Released configurations | What the figures establish | What they do not establish |
| --- | --- | --- | --- |
| Meta Newsroom, “Introducing Llama 3.1,” 23 July 2024 | 8B and 70B parameters among the released configurations | Model size and release identity | Product MAU, license status, or authorization above the contractual line |

Meta’s 23 July 2024 Llama 3.1 model card supplies the complementary first-party evidence: the released weights are linked to the Llama 3.1 Community License, and a commercial product or service crossing the agreement’s large-user boundary must obtain a separate written commercial license from Meta. The model card therefore ties the downloadable weights to license obligations; a hosting service’s “available” or “open source” label does not detach a deployment from those obligations.

The agreement is clearest when its text is not replaced by interpretation. Section 3.1, captioned “Commercial Products,” requires a separate written commercial license from Meta when a commercial product or service using the Llama Models exceeds the agreement’s large-user boundary. The defined term “Llama Models” ties the clause to Meta’s released model materials. Neither the parameter names nor a product manager’s characterization of the service modifies that operative language.

The license register should record the canonical repository, meta-llama/Llama-3.1; the full model variant; and the immutable commit used by the product. Counsel should then preserve the LICENSE and model card from that same commit alongside the production build or serving record. A repository’s live license link can move, and a hosting mirror can republish material under different assertions; neither establishes the terms attached to the exact commit the product obtained.

The quoted figure is Meta’s private contractual threshold—not an industry benchmark, statutory safe harbor, or general indication of model capacity. For approval purposes, the deploying company must substantiate its own current, license-relevant production MAU from product records. If that count exceeds the contractual line, counsel should not clear launch, expansion, or continued operation at that scale until a signed Meta commercial license covering the use is in hand. Downloading the weights again or accepting a second Community License copy does not create that distinct grant.

![8B and 70B — Meta monthly active users](https://static.mm-ais.com/article-images-pixabay/meta-monthly-active-users-community-righ-ec8157b3.jpg)

## Two License Paths

The decisive distinction is provenance, not price. On a license register, Meta’s published Llama 3.1 Community License and a separate Meta commercial license must occupy different rows: the first is the standard public grant; the second is a negotiated grant covering a particular deployment. When the current license-relevant monthly active-user count is above the Community License’s line, the separate Meta commercial license is the only compliant path. Downloading the Community License again, accepting another clickwrap, or retaining another copy does not create that grant.

| License instrument | Scale condition | Covered use | Negotiation status | Clearance document | Verdict |
| --- | --- | --- | --- | --- | --- |
| Llama 3.1 Community License | Available only at or below the license-defined MAU line and only after every other license condition is met. | Commercial use permitted by the published standard grant, subject to its model, use, attribution, and other conditions. | No deal-specific negotiation; the published terms control. | The Community License governing the model and use, supported by records showing scale and compliance with every other condition. | ELIGIBLE ONLY ON THE STANDARD PATH; not available above the MAU line. |
| Separate Meta commercial license | Required when the current license-relevant MAU count exceeds the Community License line. | The commercial product or service and uses identified in Meta’s negotiated grant. | Separate negotiation with Meta. Terms may add bespoke pricing, user scope, support, indemnity, audit, and termination provisions. | A signed Meta commercial license covering the relevant use; a quote, term sheet, email approval, or second Community License copy is insufficient. | WINNER—THE ONLY COMPLIANT PATH ABOVE THE MAU LINE. |

Scale must be a decision gate, not a proxy for economics. For a small dedicated deployment below the line, the Community License may control, but only if the model lineage, permitted use, user attribution, and remaining conditions are satisfied. A low quote does not cure a failed condition. Conversely, custom pricing does not make an above-line product Community-License compliant merely because Meta will discuss a deal. The register should therefore test the current count first, then match the resulting path to the required instrument.

The large-user path is contractually different because its terms are privately negotiated rather than supplied in the published standard grant. Pricing is only one possible term. The executed agreement may also narrow the covered user population, define support obligations, allocate indemnity, permit audits, or add termination rights. Those provisions cannot be inferred from the Community License or assumed from a sales conversation; counsel must read the signed instrument and match its product, service, users, and permitted uses to the deployment.

Unresolved model lineage, ambiguous user attribution, or an above-line count belongs in an exception lane shared with Meta and counsel—not in a default selection of the Community License. Hold launch, expansion, or continued operation at the relevant scale while the question remains open. Record the model, use, counting method, question, Meta response, and resulting document. A response can clarify the license position, but only the signed Meta commercial license changes an above-line deployment to cleared.

The controlling source for the scale ceiling and standard grant is Meta’s public Llama 3.1 Community License. For the large-user path, the signed commercial license—not pricing material or a duplicate standard license—is the controlling evidence.

![Two License Paths — Meta monthly active users](https://static.mm-ais.com/article-images-pixabay/meta-monthly-active-users-community-righ-d461d9c8.jpg)

## What the Data Doesn't Tell You

The weak point is not Meta’s license text; it is the evidence used to apply it. Public download totals, press-reported audiences, app installations, API-call volumes, and registered seats measure different populations. None, standing alone, establishes the current license-relevant monthly active-user count of a commercial Llama 3.1 deployment. A second download or clickwrap of the Community License is especially weak evidence: an acceptance event cannot supply the distinct written commercial grant required above the large-user line.

The first limitation is temporal. A dashboard can lag the product, while pilots, migrations, and enterprise rollouts can change the relevant population faster than a monthly report. The second is attribution. A direct consumer service, a white-label deployment, and an API sold through a distributor may expose different views of the same activity. Unique accounts, individual humans, downstream customer users, and automated requests are not interchangeable. Evidence close to the line therefore needs a dated source extract, a documented counting method, and a product-to-provider mapping. Neither a press estimate nor an unexplained vendor total is a dependable substitute.

Variance across cases is evidentiary, not semantic. Consider a Llama 3.1-backed help desk sold through a reseller: the reseller may report contracted customer organizations, the deployment may record individual end users, and the model gateway may see only requests. Those datasets can all be accurate while describing different populations. The strongest license-register file preserves the source query, counting assumptions, effective date, reconciliation notes, and product mapping rather than copying a dashboard headline into the record.

| Evidence item | What it establishes | What it does not establish | Counsel treatment |
| --- | --- | --- | --- |
| Public downloads or clickwrap events | Acquisition or acceptance activity | Current unique users or commercial scope | Do not treat as a scale or authorization record |
| Storefront installation figures | Distribution through that channel | Use across other channels or current activity | Reconcile against product telemetry |
| API or request logs | Traffic volume and timing | Whether traffic represents unique people rather than accounts, bots, or repeated calls | Preserve query logic and attribution rules |
| Reseller or customer console | Usage visible to that system’s controller | Downstream users, other channels, or the complete licensed population | Map the controller’s role and data boundary |
| Signed Meta commercial license | Existence of a separate commercial grant | Coverage unless the actual product, model, deployment, and use match its terms | Match the signed record to the deployment |

The large-user condition ceases to control when its legal predicates are absent—for example, when the Community License is not the governing agreement or a signed Meta commercial license already covers the use. Evidence that the scale condition is not triggered also prevents that condition from applying, although other license obligations may remain. But uncertainty does not suspend the rule. If the current count may exceed the line, counsel should not approve launch, expansion, or continued operation at that scale until the signed Meta grant covers the use. Imperfect telemetry creates a verification obligation; it does not create an exception.

![What the Data Doesn't Tell You — Meta monthly active users](https://static.mm-ais.com/article-images-pixabay/meta-monthly-active-users-community-righ-1a381fba.jpg)

## The Large-User Line Without a Counting Recipe

At the large-user line, monthly active users (MAU) are a legal switch, not a price quote. The public trigger can be undisputed while the population behind it remains contestable. Counsel therefore needs an auditable chain from telemetry definition to product, feature, checkpoint, and contract; scale evidence alone cannot establish what Meta licensed or whether the required commercial grant is already in hand.

| Unresolved Point | Evidence and Record | Legal Consequence |
| --- | --- | --- |
| Commercial economics | The Community License supplies the large-user trigger but publishes no standard price, minimum commitment, term, service level, indemnity, or termination formula for the separate Meta deal. | Treat the user count as a gating fact, not as a valuation for the required written commercial license. |
| Meaning of MAU | “Monthly active users” does not settle whether to include direct model users, everyone served by an embedded feature, API customers, duplicate identities, or account aliases. | Document a defensible telemetry interpretation, including the population, activity event, deduplication method, measurement period, and accountable owner. |
| Corporate and product scope | A parent company’s consolidated MAU does not automatically equal the MAU of every product or service. Counting only direct model sessions may omit users covered by a broader product-level reading. | Preserve product-level and feature-level records, together with the roll-up methodology used for the licensing decision. |
| Counter-evidence from scale | According to Meta’s Q2 2025 earnings call, Meta AI had passed 1 billion MAU. That figure does not identify the exact model version, license-relevant population, or third-party licensing obligation. | Do not enter it as proof that the large-user clause was triggered for a commercial Llama 3.1 deployment. |
| Spikes and corporate events | The public license materials supply no cure period or waiver standard for a one-month spike, acquisition, retroactive authorization, or product reorganization. | Pause the affected decision and obtain legal review rather than inferring a grace period from silence. |
| Checkpoint drift | The 2024 Llama 3.1 text does not set the terms for later checkpoints. | Audit the exact model, contract version, acceptable-use policy, and signed agreement; Llama 3.2, Llama 3.3, or Llama 4 may carry different terms. |

If the current license-relevant MAU is above the large-user line, do not approve launch, expansion, or continued operation at that scale until counsel holds a signed Meta commercial license covering the use. The approval record should identify the telemetry interpretation, product and feature population, exact checkpoint, governing contract, and signed grant. An unresolved counting question is not permission, and a second download or clickwrap of the Community License cannot cure the hold.

![The Large-User Line Without a Counting Recipe — Meta monthly active users](https://static.mm-ais.com/article-images-pixabay/meta-monthly-active-users-community-righ-03ed2798.jpg)

## 9M to the Large-User Line: A Complete Llama 3.1 Boundary Case

HarborHelp’s boundary failure concerns scale, not model stewardship. HarborHelp is a hypothetical customer-support service running Meta’s Llama 3.1 70B model. Meta’s public model materials identify the 70B checkpoint, and Meta’s Community License supplies the applicable large-user line; HarborHelp’s monthly counts are constructed for this calculation, not presented as Meta telemetry. That distinction keeps the evidence auditable: the decisive fact is the service’s current deduplicated monthly active-user count, not its traffic or parameter count.

| Period | Deduplicated evidence | Boundary calculation | License disposition |
| --- | --- | --- | --- |
| January | A deduplicated user count below the license-defined line, with each counted user invoking the AI support feature at least once that month | Below the line, with remaining headroom | Below this specific clause; the scale condition does not bar this deployment |
| February | A deduplicated monthly-user count above the license-defined line after the AI feature becomes the default | Above the line and higher than January | Do not approve launch, expansion, or continued operation at that scale until counsel holds a signed Meta commercial license covering the use |

The February result does not change because support messages multiply or because the deployed checkpoint remains the 70B model. Neither fact creates another active user, and neither changes the contractual MAU measure. HarborHelp therefore crosses the line through its deduplicated user population, even though its model and message volume remain constant.

The scope record should identify the verified 70B model hash, unmodified weights, the single named HarborHelp service, the absence of model redistribution, and HarborHelp as the contracting product owner. Those entries resolve the model and deployment perimeter cleanly. They do not resolve February’s scale condition: a tightly controlled model scope cannot supply the separate commercial permission that the public Community License does not grant above this threshold.

The February license-register entry should state exactly: *“Community grant—separate commercial license required.”* It should also name the telemetry owner responsible for the count and preserve an evidence link supporting the deduplicated February result. The rollout decision should be recorded as blocked. Neither another clickwrap nor a second copy of the Community License is the distinct written commercial grant Meta requires.

HarborHelp should pause expansion and seek written commercial terms addressing the already-triggered month, including express retroactive authorization where necessary. Counsel should record consideration as *“to be negotiated”* because Meta’s public license states no fee; that silence does not establish permission to operate at scale. No launch, expansion, or continued operation at the recorded February scale should be approved unless and until counsel holds a signed Meta commercial license covering that use.

## Count, Scope, Contract

Read Meta’s public Llama 3.1 Community License Agreement as the source of the numerical boundary, then implement it through a product-level license register. Clearance is a reconciliation problem, not a download problem: the commercial grant must match the population, model artifact, delivery path, product surface, scale, term, and territory. A second Community License download or clickwrap adds no commercial grant.

| Control | Register test | Required disposition |
| --- | --- | --- |
| Rule 1—Count | For each commercial product or service using Llama 3.1, calculate unique monthly users under the agreement’s wording. If attribution is disputed, use a conservative upper bound. Do not substitute downloads, parameters, tokens, chats, or concurrent sessions. | Preserve the count’s source, assumptions, exclusions, and reviewer. |
| Rule 2—Boundary | Treat a count at the license-defined line as not over it and a count above the line as over it. | Route an above-line month to legal clearance before approving that scale. |
| Rule 3—Scope | Before measuring MAU, record the exact checkpoint or hash, fine-tune or derivative status, direct/API/embedded path, contracting entity, and product surface. | Use those fields to identify whose count must be reconciled with whose grant. |
| Rule 4—Governance | Set an internal amber escalation below the numerical line and the first over-line count as the red control on new scale. | Label both as company controls, not license terms. |
| Rule 5—Proof | Accept only a signed Meta commercial license covering the model, products, user scale, term, territory, and permitted deployment. | Reject another Community License acceptance, reseller assurance, or counsel memo alone. |

Keep the counting record as an evidence packet, not a dashboard tile. Preserve the monthly window, source extracts, agreement-based deduplication method, exclusions, attribution disputes, and rationale for any conservative upper bound. If a person could reasonably be assigned to more than one product surface, retain the unresolved attribution and its effect on the result. A total without that lineage cannot support a defensible clearance decision.

Fix scope before measuring because changing any key can change both the relevant population and the required license match. For example, a signed Meta license covering a named API product does not, by itself, clear an embedded assistant or a fine-tuned derivative offered by a different contracting entity merely because both began from the same checkpoint. Record the artifact, delivery path, entity, and surface first; calculate MAU against that defined scope.

Before an approval ticket advances, require the product owner to deliver the count packet, engineering to attest the scope, counsel to match it against the signed Meta license, and the approver to record the result. If the count is above the line or any licensed-scope field does not match, hold launch, expansion, and continued operation at that scale until counsel holds a signed Meta commercial license covering the use. A memo may explain the gap; it cannot create the missing grant.

## What to do next

| Step | Action | Why it matters |
| --- | --- | --- |
| 1 | For a 2026 commercial Llama 3.1 deploym Frequently Asked Questions Does a commercial product with exactly 700,000,001 monthly active users remain eligible under the Community License? No; 700,000,001 is above the applicable MAU line, so the Community terms cease to provide the relevant license and a separate Meta commercial grant is required. What must counsel hold before approving an above-line launch, expansion, or continued operation? Counsel must hold a signed separate Meta commercial license covering the relevant use. Can another download, clickwrap acceptance, or Community License copy substitute for Meta’s commercial grant? No; the Community License and Meta’s separate commercial license are distinct legal instruments, and a download receipt does not substitute for the commercial grant. Which product metric determines whether the large-user clause applies? The determining metric is the commercial product or service’s current license-relevant monthly active-user count, not parameter count, tokens, API calls, conversations, or concurrent sessions. Do the 8B and 70B model sizes affect the MAU threshold? No; 8B and 70B are versioning facts, while the large-user boundary is determined contractually by monthly active users. Is there a standard public application process or fee for the commercial license required above the MAU line? No; the supplied material discloses neither a standard public application form nor a fee for Meta’s separate commercial grant. Quick answers Does 700,000,001 monthly active users trigger the Llama 3.1 large-user clause? | Yes—the trigger is a strict inequality, so the next monthly active user above the license-defined line is inside the clause. |
| What document is required above the applicable MAU line? | A separate written commercial license signed by Meta is required. |  |
| Does the MAU threshold measure model size or technical capacity? | No—the large-user MAU threshold is a contractual escalation point, not a model-size test, public price tier, or capability benchmark. |  |
| Has Meta disclosed a standard public application form or fee for the commercial grant? | No—the supplied material identifies no standard public application form or fee for the separate Meta commercial grant. |  |
| What must counsel do before approving a deployment above the contractual MAU line? | Counsel must obtain and hold a signed separate Meta commercial license covering the use before approving launch, expansion, or continued operation at that scale. |  |

Also worth reading: **Patent Renewal Fees 2026: $7,700 Prune vs Renew Decision**: [Patent Renewal Fees 2026: $7,700](https://iprs.cloud/blog/patent-renewal-fees-2026-7700-prune-vs-renew-decision.php) · **2026 Centralized License Register: Legal Review 5 Days to 1**: [2026 Centralized License Register: Legal](https://iprs.cloud/blog/2026-centralized-license-register-legal-review-5-days-to-1.php) · **API Triangulation and Clearance Velocity Data for IP Teams**: [API Triangulation and Clearance Velocity](https://iprs.cloud/blog/api-triangulation-and-clearance-velocity-data-for-ip-teams.php)

### Related reading

- [Can Prompts Qualify for Copyright?: Section 202.1 in 2026 — Length or Human Expression?](https://iprs.cloud/blog/can-prompts-qualify-for-copyright-section-2021-in-2026-length-or-human-expression.php)
- [Trade secret disclosure rules: $407M verdict on binder vs clean team](https://iprs.cloud/blog/trade-secret-disclosure-rules-407m-verdict-on-binder-vs-clean-team.php)
- [Patent Maintenance Fees: 30% Lapse, Auto-Pay vs Triage vs Prune](https://iprs.cloud/blog/patent-maintenance-fees-30-lapse-auto-pay-vs-triage-vs-prune.php)
- [Using Li Brocade Patterns for 2026: Pay 5% vs Hold for Redesign](https://iprs.cloud/blog/using-li-brocade-patterns-for-2026-pay-5-vs-hold-for-redesign.php)
- [WIPO GREEN Register: 2,100 Self-Declared Connections, Not Vetted](https://iprs.cloud/blog/wipo-green-register-2100-self-declared-connections-not-vetted.php)
- [Anaqua AQX 14-Day AI Docket: 74% of 1,067 at GPL Risk](https://iprs.cloud/blog/anaqua-aqx-14-day-ai-docket-74-of-1067-at-gpl-risk.php)

### Latest

- [Can Prompts Qualify for Copyright?: Section 202.1 in 2026 — Length or Human...](https://iprs.cloud/blog/can-prompts-qualify-for-copyright-section-2021-in-2026-length-or-human-expression.php)
- [Trade secret disclosure rules: $407M verdict on binder vs clean team](https://iprs.cloud/blog/trade-secret-disclosure-rules-407m-verdict-on-binder-vs-clean-team.php)
- [Patent Maintenance Fees: 30% Lapse, Auto-Pay vs Triage vs Prune](https://iprs.cloud/blog/patent-maintenance-fees-30-lapse-auto-pay-vs-triage-vs-prune.php)

Canonical: https://iprs.cloud/blog/meta-monthly-active-users-community-rights-end-at-700000001.php
Markdown: https://iprs.cloud/blog/meta-monthly-active-users-community-rights-end-at-700000001.php/index.md