A product team is about to ship a Llama-4 fine-tune as part of a consumer feature inside an app that has had more than 800 million monthly active users since before Llama 4 was released. They've assumed the Llama community license is "basically open" and have not requested anything from Meta. Identify the specific clause they need to re-read, what threshold it sets, why it applies to a fine-tuned derivative as much as to the base model, and what they need to do BEFORE shipping.
The Llama Community License requires a separate Meta license above 700 million MAU across the licensee's products at release. Fine-tuning does not reset the clock: the derivative inherits the same threshold.
Imagine borrowing a recipe from a chef who lets anyone use it, except for restaurants serving more than seven hundred million meals a month. Those huge restaurants have to call the chef first and arrange a separate deal before they can serve any dish that uses the recipe. Tweaking the recipe with extra spices does not change anything, because the new dish still uses the original. The size of the restaurant you serve in is what triggers the call, not whether the recipe section of the menu is busy. So if your restaurant chain has over seven hundred million monthly customers, you make the call before you open the kitchen.
Concept explanation~2 min read
Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.
Concept explanation~2 min read
Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.
The Llama Community License sits in an unusual place in the model-licensing landscape. It is far more permissive than a closed-API arrangement. That is why teams reach for Llama bases when they want to control fine-tuning, hosting, or fine-grained customisation. But it is not a permissive open-source licence, and conflating it with Apache 2 or MIT is a costly licensing mistake.
The gap between actually open and basically open shows up in one specific clause. The Llama Community License contains a monthly active user threshold, set at 700 million. Above that threshold the community licence ceases to grant rights. A separate commercial licence from Meta is then required before any use of the model or its derivatives. The clause was designed to catch this case: a large consumer app reaching for an open-weight model without negotiating a deal first.
What the clause says
The relevant section of the Llama Community License is the additional commercial terms. The language, paraphrased and reduced to essentials, is as follows. Suppose that on the day of the Llama version release, the monthly active users of the products or services made available by you or your affiliates exceed 700 million in the preceding calendar month. Then you must request a licence from Meta, which Meta may grant to you in its sole discretion. You are not authorised to exercise any of the rights under the agreement unless Meta otherwise expressly grants you such rights.
Four things stand out. First, the threshold is a hard numeric boundary, 700 million MAU, not a range or a soft cap. Second, the clause uses sole discretion, meaning Meta is not obliged to grant the licence on request. Third, the clause covers the licensee and its affiliates. That closes a structural loophole. A large parent company might try to route Llama usage through a small AI subsidiary to argue the subsidiary alone is under the threshold. Fourth, the clause is keyed to the day of release of the Llama version you use, with the MAU measured in the preceding calendar month.
The practical reading for this team: the app already had more than 800 million MAU when Llama 4 was released. So the company is over the threshold before any AI feature ships. The licence does not grant rights to use the Llama-4 base or any derivative of it. The team is operating without a licence unless and until Meta grants the commercial one.
Situations where this technique stops working.
2–4 min · Everything important, quickly.
Real products, models, and research that use this idea.
- Meta released Llama 4 Scout and Maverick in April 2025 under the Llama 4 Community License, with the same 700 million MAU threshold structure as Llama 2 and Llama 3.
- Mistral's Apache 2 licensed open-weight models like Mistral Small 3 carry no MAU cap, so they are a natural swap target for large-MAU products.
What an interviewer would ask next. Try answering before peeking at the approach.
QHow would you advise a company whose MAU is currently 500 million but projected to cross 700 million within the next year?
The test is measured in the month before each Llama version's release date. A version released while the company is at 500 million stays under the community terms even after later growth, but any version released after the company passes 700 million needs Meta's licence. Start the commercial-licence conversation before you plan to move to a newer version, and keep a permissive-licence base ready as a fallback.
Red flags & common mistakes
The phrases that signal junior thinking. Click to expand.
Red flags & common mistakes
The phrases that signal junior thinking. Click to expand.
Believing that fine-tuning produces a fresh license-free model. The derivative inherits the upstream terms, including the MAU threshold. The LoRA training step changes nothing about that.
60 second bullets to scan on the way to the call.
The exact numeric MAU threshold in the Llama Community License
What the threshold counts (all products of the licensee and its affiliates) and why
Primary sources. Browse if you want the original framing.
Same topic, related formats. Practice these next.