Apparently one of the biggest questions people are asking online is about the difference between the three tiers Meta offers Muse users. In general, the creators of AI agents are bad at explaining this across the board, not just Meta. Here are the tiers explained simply.
1. The Free plan
Meta describes this as "standard use and light task delegation," and they expect most people to stay here. OK, this is the basic plan you have immediately after you download a Muse agent. Your Muse agent has no name, no personality or anything. It's a blank slate. You don't need to spend any money to use the agent and you don't need to add credits or pay for anything at all. Downloading the app and using the Free tier is just that. It's free.
The hurdle is the agent has to be linked to your Facebook or Instagram app. Basically, the Muse agent goes to your Meta accounts center, and links to the visible accounts in there. If you are on Facebook, Instagram, and Threads, it will link to all three and verify your identity that way. This happens so fast, it's seamless. But before you sign up, think about whether you're OK with giving Meta more of your data.
On one hand, I've spent most of my time with AI building local AI agents that are completely private. The world does not know they exist. But on the other hand, I'm not a developer or a programmer. I don't even work in tech. And the Free tier of Muse is better than the 6 agents I spent weeks building. It was a good learning experience, and I appreciate that they are still private. But Muse is still better.
This tier comes with 100,000,000 context tokens per week, or roughly 400,000,000 per month assuming there are exactly four weeks in a month. If you are not a heavy AI user, this is probably plenty. If you downloaded Muse and you're just testing it out as a hobbyist and you don't actually have a plan for what your Muse agent is going to do — this is more than enough. You can send yourself daily news articles the agent wrote that are unbiased. You can have it plan out your day in advance after looking at your Google calendar, you can have it plan a yoga routine or a vacation, or a holiday party or a date night or something. You can probably do all of these things in one seven-day period on the free plan. If you're a hobbyist, and you spend the first half of the week just chatting and you aren't using the bot in an agentic way (having it go do things for you on its own and then report back in 5 hours) you might not even run out.
2. The Power plan
Meta describes this as "more usage for handing off everyday tasks." This is the plan I'm using. This gets you 500,000,000 tokens per week or 2,000,000,000 per month assuming there are exactly four weeks in a month. Unless you have a deeper understanding of coding, this is probably plenty for 85% of the Muse users in the world. The budget breaks out to about 71,428,571 tokens per day (500,000,000 ÷ 7 days).
The trouble with non-techies or normies like me is we don't actually know how much usage that is. How does it compare to Claude or ChatGPT or something like that? We really don't know. It's still early. What I do know is even as a heavy user sending several messages every hour, I don't use 71,000,000 context tokens per day.
When I realized how helpful the free plan was and thought through the fact that the "power" plan could probably save me more money per month than it's costing me, I decided to go ahead and upgrade. I don't know your budget, but if you're an adult with enough money to sustain the lifestyle you're currently living, you can probably have your agent's first project be to find a way to cancel old subscriptions, find discounts on food and gas, and find a way to cover the $20 a month it's costing you. I have no doubt that Muse could do it. I'm not saying it would be easy, or fast. But you could do it and have tokens left over. This really comes down to how frugal you already are also. But if your Muse agent's number one priority is to save you $30 a month and your income and expenses are high, or you have numerous subscription services, I have no doubt Muse could do it. I've even heard of people having Muse wait on hold for them with utility providers, or haggle with online chat bots to negotiate the cost of various goods and services down.
If you want to get organized, if you want to embrace AI, or if you have a business and you want to take a step forward into the future, take this step and pay the $20. If you are completely against paying $20 a month to use Muse, you should either stay on the free version as long as Meta will allow it, or invite a friend on Facebook or Instagram. If you choose to invite a friend, there are some nuances about that process. Basically, they have to accept your invitation, and then go into their own settings within 48 hours and press the button to "redeem," acknowledge or accept that you were the one who suggested they join.
The good news here is Muse gives you 1 billion context tokens for free each time you do this. The only caveats are your friends have to download the app and redeem your invitation bonus within 48 hours or it disappears. 1 billion context tokens is equal to ten free weeks of context. They don't ration your context or hold it from you, they just keep it sitting separately at the bottom of your usage meter. Once your 100 million free context tokens for the week run out on the free plan, or once your 500 million context tokens run out for the week on the Power plan, your Muse agent will automatically start using the free bonus you were awarded for your friend redeeming your invitation bonus. It's not just 1 billion context tokens for you — your friend gets this bonus too.
Take a look at my old post where I share my own invitation. If you have never used the Meta app before, this is a great way to start with 1.1 billion context tokens for free right when you download the app. You'll get the 100,000,000 weekly context tokens just for having the app, and the 1,000,000,000 bonus tokens that do not expire from going into your settings and acknowledging that I shared my link with you. There is probably a YouTube video that explains this more clearly somewhere. If you're still confused, try to find that.
As of the time of this writing, my Muse agent does not know how many tokens I have, and can't see that. I tried to ask him, "Did they go into their settings and accept my 'redeem' coupon and actually activate the 1 billion tokens for me and for them?" But Max does not have the ability to see that information. You can go into your settings and see how much context you have already used for the week, what day and time it resets, and what percent of your total context you have left. It's relatively easy to monitor your own usage manually.
I have my Muse agent Max doing more than 10 cron jobs every day, conducting research tasks and working on light tasks throughout the day. More often than not, I don't use the full daily allotment if you break it out by day. It's legitimately hard to use all of the tokens before the end of the week unless you're using it all day every day.
I told my Muse agent "look Max, we've had 5 people sign up through my affiliate link, so that's 5 billion context tokens and we're barely using the full 500,000,000 context tokens for the $20/month plan now. So let's just use a ton of tokens to test this technology." In order to test the token burn, I told my Muse agent to scrape the web and find the top commercial real estate sales brokers in every state, add them to a Google sheet with their name, phone number, email address, and a link to their LinkedIn profile. Then upload it to Google Drive when you get a decent amount of names on there. I could still ask Muse plenty of questions while he was doing that in the background. Admittedly, this burned more tokens per day than the daily allotment to have that running. After 2-3 days, I told him to slow down on the web scrape and just scrape the web for 3 hours a night in the middle of the night, and add the new names as you go, but stop after 3 hours to conserve tokens. I also told him to slow down if we get to less than 10% context tokens for the week which he happily did. By the end of the week I had over 1,000 brokers' names and contact information on the sheet, but I still had 12% of my context tokens. Then I really cranked it up and tried to just burn them on random tasks to see how long it would take to run out. By then I had already accumulated my referral bonus. So I knew if it ran out, I could just use some of the extra tokens. But even after heavy usage, Max didn't run completely out of tokens until later in the day on the last day. So that was pretty wild.
Unless you have some enterprise use case, I can't imagine why you would need the Max plan, as I was able to get so much done with the basic plan. But again I'm a normie, so who knows.
3. The Maximum plan
The "Maximum" plan or the "Max" plan is the $100/month plan. Meta describes this as "the highest usage limits for heavy delegation." This gives you three billion context tokens per week, or 12 billion tokens per month. Meta keeps this ambiguous for some reason, but I did the research so you don't have to. It's 3 billion context tokens per week. That's enough to get a lot of shit done. I'm betting if programmed properly, your Muse agent will be more productive on the computer than you are for that many context tokens.
There are several misconceptions about the Max plan I want to clear up now though. For starters, based on what I've read, this is not going to make your Muse agent any faster. Meta context tokens do not equal speed. It's not giving you a better agent, or a smarter or faster agent. It is giving your agent more runway during the 7 day period. So if you think of it like regular gas in a car on the free plan, the power plan is not higher octane fuel and it's not a different bot or a smarter model. The power plan is superior to the free plan because your gas tank in your Muse car on the free plan only fills half way up. The Power plan fills your car up to a full tank. The "maximum" plan is like filling up your gas tank for a road trip and getting some gas canisters and filling those up too and putting those in the trunk. You can go a farther distance or have your agent running more agentic tasks in the same 7 days. But he will not do those tasks any better or faster than someone using the free plan. The Max plan will just be able to go farther in the same 7 days without needing to stop at a gas station.
I've said it before on this blog and I'll say it again, for most people, the way to get as much usage as possible is to invite friends to join you and get the 1 billion context redeem reward, or if you don't have the Meta app yet, click on my link in my other blog post with my code IKSL9N and after you click "redeem" in your settings, you will get an extra 1 billion context tokens up front. Even if your computer has more RAM, paying more for a maximum plan will not make your Muse agent operate faster.
Let's talk about real world examples where people have actually had to use the Max plan to get something done. Here are a few examples. I read one story about a developer using Claude Code with MCP servers to automate his workflow who kept hitting the limit in Claude Code. So he got the max plan to have his Muse agent assist to keep that whole process going. Another user said the way Meta is doing the math on all this, if you're paying $100 a month for max Muse usage and you are a legitimate business, it's going to be very generous how much they will allow you to get done. That guy said he only got past 80% of his weekly usage one time in one especially busy week since Muse came out. Muse is still so new. I'm sure someone is hitting the limit somewhere. Finally one third user mentioned that Max is for business use and the only individual who might need that much context is if you're running numerous cron jobs and monitoring multiple subagents around the clock 24/7. So yeah. It's a lot of usage.
Lastly: the math nobody does for you
Here's another way the AI community could start thinking about the context for Meta usage. Most people who bother paying at all are probably paying the $20/month plan. It's right in the middle of their selection, you get more context than the free plan, while still saving a bit of money rather than getting the maximum plan. So for most paying Meta users, $20 per month comes to $0.67 per day. So for those of us on the Power plan, we are paying approximately $0.67 per day for 66,666,666 context tokens per day (2,000,000,000 ÷ 30 days). This is the consensus that most paying Muse members have agreed the context is worth. Most of us who feel as though it's worth paying anything at all are in agreement that if we're going to pay, we're willing to pay $0.67 per day for 66,666,666 tokens per day of usage.
With ChatGPT, if you're in the middle of a project and you run out of context, you can add a minimum of $10 and Codex will finish what it was working on. Meta doesn't work that way though. It's either free, or $20, or $100 and nothing in between. You can't buy context tokens in $1 increments. I'm just saying for the sake of argument if you could, $1 equals roughly one hundred million tokens. 1 penny gets you one million context tokens. Or think of it this way. If you want to buy in bulk and we assume a barrel of oil is $100/barrel this week, one barrel of oil at $100 buys you 12 billion context tokens. One Satoshi of bitcoin gives you 84,000 context tokens.
These numbers will change over time for several reasons, mostly because Meta is subsidizing the cost of AI usage the same way ChatGPT and Claude are. These companies are giving you very cheap AI usage now, in hopes that you'll get hooked and then they can charge more later or someone will come up with a way to make it more affordable later. The most likely outcome is probably that as the price of power rises and each of these platforms has more people using the tools every day, the price of the tools will rise. It's costing Meta more than $20 per month for me to use my Muse agent heavily every day. They are paying the extra portion. Based on Meta's pricing model, the free Muse agent users are effectively getting $4 per week of usage at today's usage prices... for free. Thanks Meta. I'll take free while I can.
I'm calling it out now though guys. When the cost of power rises over time, eventually the free tier will be $5, and the $20 tier will be like $50, and that's just the beginning. So be careful how heavily you rely on this technology. Or do what I'm doing and try to get it all set up now while usage is cheap. LOL
Quick answers
What do the three Muse plans cost?
Free is $0, Power is $20/month, and Maximum is $100/month. There is nothing in between — Meta doesn't sell context tokens in smaller increments.
How many tokens does each plan get per week?
Free: 100 million context tokens per week. Power: 500 million per week. Maximum: 3 billion per week. These are the best-supported figures from independent tracking; Meta's own wording is less precise.
Does the Maximum plan make Muse faster or smarter?
No. It buys more runway in the same 7-day period — not a faster agent and not a smarter model. Same gas, bigger tank.
How does the invite bonus work?
Your friend accepts your invite and hits "redeem" in Settings within 48 hours. You each get 1 billion tokens, sitting separately at the bottom of your usage meter until your weekly allowance runs out.