Muse Code subscriptions cost $5 to $200 a month. If you mainly use Muse Spark 1.3, compare the subscription fee with what the same workload would cost in the same mode on the Meta Model API. If you need Muse Wonder Full, Power Usage is the cheapest way in at $50 a month. Judge it by how many tasks it completes to your standards within the two quota windows and how much manual work it saves.
What you get with each Muse Code plan
| Plan | Monthly price | Advertised allowance | Muse Wonder Full |
|---|---|---|---|
| Everyday Usage | $5 | About 10–50 prompts every five hours | No |
| High Usage | $15 | 5× Everyday | No |
| Power Usage | $50 | 20× Everyday | Yes |
| Maximum Usage | $200 | 80× Everyday | Yes |
Every plan has a five-hour window and a weekly window. One prompt can lead to file reads, multiple model calls, and fixes after tests. The number of projects you can finish depends on how much work each prompt sets in motion. Pricing and availability may vary by region. Meta’s plan details · Compare Muse Code plans.
How much Power quota Muse Wonder Full uses
In October, Kingy measured Power Usage on Muse Code 1.4.7. The main agent used Standard mode at Medium reasoning effort. Kingy recorded the change in both usage meters for each task. Kingy’s tests.
| Muse Wonder Full task | Time | Five-hour quota used | Weekly quota used |
|---|---|---|---|
| Pagination | 1 min 43 sec | 3 percentage points | 1 percentage point |
| Invoice processing | 2 min 41 sec | 4 percentage points | 1 percentage point |
| Tracker | 9 min 51 sec | 15 percentage points | 5 percentage points |
A meter moving from 33% to 36% means the run used three percentage points of that allowance. Both meters show whole numbers, so readings for short jobs are approximate. Both Wonder variants passed the external acceptance checks. For small utilities, the extra cost of Full makes sense if it improves the share of results you can accept on your own tasks.
In a longer VulcanBench run with Muse Wonder Full in Contributor mode, the weekly allowance ran out on October 14 after 23 Low-effort tasks and 21 completed Medium-effort tasks. The run resumed on Muse Code 1.4.9 after a quota reset. On November 3–4, the same version ran 23 Max-effort tasks. For long batches, keep an eye on what’s left for the week after the five-hour allowance refreshes. VulcanBench report.
How to read the Spark 1.3 value estimates
Two Everyday subscribers converted weekly use into very different monthly values. A workload with a high cache hit rate came to about $3 at Contributor API prices. The second subscriber used 98% of the weekly quota and estimated about $20 a month for calls tied directly to their tasks, or around $26 including Muse Code’s own reminders and approval checks. First measurement · Second measurement.
On the $15 High plan, an experiment logged Spark 1.3 calls and changes in the weekly usage meter in separate runs. The full-month estimates were about $140 at Standard API rates and $87 at Contributor API rates. The experiment measured part of a week, so the value you get depends on how much of the allowance you use and how much of your input is served from cache. Experiment logs.
When to upgrade to Power or Maximum
Power costs $35 a month more than High. If you regularly need Full, pick a few recent jobs where regular Wonder left you with edits to make, and try similar jobs with Full. Compare the share of accepted results, the time you spend on manual fixes, and both quota counters before and after each run.
Maximum costs $150 a month more than Power. It includes the same model and four times the advertised allowance. Judge the upgrade by the extra work you can finish in a month. If Maximum lets you complete 20 additional tasks, the upgrade costs $7.50 per extra task. Use your weekly Power usage and the time you spend waiting for a reset as a starting point.
If regular Muse Wonder meets your needs, you can pay per token through the Meta Model API. The rates are $4 per million uncached input tokens, $0.20 per million cached input tokens, and $20 per million output tokens. At those rates, 5 million uncached input tokens, 50 million cached input tokens, and 1 million output tokens cost $50. Muse Wonder Full has no separate per-token price. Meta Model API pricing.
Frequently asked questions
How do I check usage and reset times?
Run muse usage to see your subscription tier, usage in both windows, and reset times without making a model call. Record the counters before and after a run within the same window. Other agents on the account can also move the shared counters. Muse Code changelog.
What changes when I turn off Contributor mode?
Contributor mode is on by default. Meta offers more usage in exchange for permission to train on submitted prompts, code, files, context, and responses. When you turn it off in settings, future content is excluded from model training. Muse Code saves the setting separately on each computer. Subscription terms.
Does Max reasoning in the API give me Wonder Full?
The max setting changes the depth of reasoning for regular Muse Wonder. To select Wonder Full, use Muse Code with a Power Usage or Maximum Usage subscription.