Who measures this
Updated 11 November 2026
One person, trading as WattMate, a sole proprietorship in Poland. There is no team and no agency: one person writes the code, runs the measurements, writes every page and answers for every number. The full seller details are on the Terms page.
Almost every article here answers with a number a question the internet usually answers with adjectives ("Chrome is a battery hog", "quitting apps doesn't help"). A number is only worth the method behind it, so the method and the machine are printed next to it. How we count says what the app measures and where the measuring stops.
Our lab: the Macs the numbers come from
| Mac | macOS | Battery: cycle count, Maximum Capacity |
|---|---|---|
| 14-inch MacBook Pro 2023, M3 Max, 36 GB | 27.0 | 652 cycles, 85% |
| 14-inch MacBook Pro 2021, M1 Pro, 16 GB | 27.0.1 | 1,730 cycles, 75% |
| 13-inch MacBook Pro 2020, M1, 16 GB | 26.7.1 | 540 cycles, 83% |
| 14-inch MacBook Pro 2021, M1 Pro, 16 GB | 14.2.1 | 1,052 cycles, 81% |
| 13-inch MacBook Air, M5, 16 GB | 27.0.1 | 12 cycles, 100% |
| 13-inch MacBook Air, M2, 16 GB | 27.2 | 195 cycles, 88% |
All 6 ran the checks of WattMate's own numbers: the app against each Mac's own power sensor, and that sensor against a USB-C inline meter not made by Apple (ChargerLAB POWER-Z KM003C), with a smart plug at the wall as a third, coarse instrument. Rules, thresholds and scripts were published before the first counted run. On all 6 Macs, WattMate listed our test programs first, and Apple's sensor stayed within about 0.3 W of the meter (the results).
The app pages also use the everyday log of one more Mac, a 14-inch MacBook Pro with M1 Pro on macOS 26.3.
The Macs differ a lot, and so does their use. In everyday logs the quiet M1 Pro on macOS 26.3 drew 8.5 W in an average window, the busy M3 Max 27.2 W, so "how many watts does my MacBook use" has no single answer. The M1 Pro on macOS 14 reported 132 energy counters and the M3 Max on macOS 15 reported 285, under different names: the set of counters depends on the chip and the macOS version. Each page names the Macs behind its numbers.
What the lab has measured
Each study names its Macs, its method and its limits on its own page.
- Does AI drain a Mac battery? ChatGPT, coding agents and a local model, measured against each Mac's own power sensor. Data and scripts are on Zenodo.
- Why 2 short messages can use so much of an AI agent's limit: token logs and 2 tests in Claude Code, with the rules of the pause test written down before its data.
- Checks of WattMate's own numbers: the app against each Mac's sensor and an inline meter, with the rules published before the first counted run.
- Who wakes your Mac at night: the silent wakes macOS records with the lid closed, and the services behind the scheduled ones.
- Why about half of a Mac's CPU energy has no owner: what per-process counters miss.
- Does quitting apps save battery? Measured: when quitting buys almost nothing, and when it buys real minutes.
- How many watts a MacBook uses, and why there is no single number.
The instruments, by name
Nothing here is estimated from CPU percentages. The readings come from sources macOS and Apple silicon already expose, several of them interfaces Apple does not document, named so you can go and read them yourself.
| Instrument | What it gives us |
|---|---|
| System management controller | The whole Mac's power, once a second |
libIOReport | The chip's own energy counters: processor, graphics, Neural Engine and a few hardware blocks |
ri_energy_nj | Each process's processor energy as the kernel counts it, the basis of each app's estimate |
| Resource coalitions | The energy of short-lived helper processes, credited back to the app that spawned them instead of vanishing when they exit |
pmset -g log | The power log macOS already keeps, which records sleeps and wakes and, for the scheduled wakes, the service that asked |
AppleSmartBattery | The battery's own counters: charge, capacity, cycle count, voltage and current |
What each of these can and cannot answer (including the parts that stay unmeasurable, like separating the idle floor from an app's draw) is set out in How we count. How the per-process counter becomes one row per app is on Watts per app, and why about half of a Mac's CPU energy has no owner is on a page of its own.
Data and code
| Record | What is in it | Where |
|---|---|---|
| AI battery study | 1 CSV row for each number in the study's results, a table per measurement series, and scripts, standard library Python only, that recompute most of those rows and compare them with the results file. Data is CC BY 4.0, code is MIT | doi.org/10.5281/zenodo.23391844, with a copy on GitHub; its scripts recompute 2,822 of the 2,904 rows |
| Checks of WattMate's numbers: rules | The rules of the checks, with their scripts frozen by hash | doi.org/10.5281/zenodo.23697150 |
| Checks of WattMate's numbers: results | The results and the data of every counted run, added to the same record | doi.org/10.5281/zenodo.24002711 |
The other 82 rows of the study compare WattMate with the lab's instruments and need WattMate's own database, which records every app on the Mac and is not published, or lab logs that are not in the archive (the check). If a number from these studies looks wrong, these archives are where you can check it, and support@wattmateapp.com is where you can say so.
The rules every page here follows
- Every number says where it comes from, and whether it was measured or worked out. The whole Mac's power, the chip's energy and the battery's counters are measured; a row per app is an estimate built from them, and the pages say so. Each article names the machine and how many measurement windows are behind the figure. If a number came from one Mac, the page says one Mac.
- Where something can't be measured honestly, the page says so instead of passing an estimate off as a measurement. Several pages end with a section on what the method cannot reach.
- Facts about other apps are checked against the source, with a date. Comparison pages carry a disclosure line saying WattMate is our app. Where the honest answer is a competitor, the page says that too: the AlDente page opens by explaining when you should keep using AlDente.
- Dates change when facts change, not to look fresh.
When a number turns out to be wrong
It has happened, and we correct it on the page where it stood.
What money does and does not buy here
WattMate is sold from this site and, as a separate edition, on the Mac App Store, and that is the whole business model: no advertising, no sponsored placements, and nothing on this site is paid for by anyone else. No page here has been written in exchange for a review, a listing or a link, and none will be.
Getting in touch
Corrections are welcome. If a number on this site is wrong, say so and it gets rechecked: support@wattmateapp.com.
Journalists and researchers: the raw analysis behind the night-wake study, method and limits included, is on its own page, and the guides are indexed at Mac battery guides.
Related: how we count · how we check · press · all the guides
Measure your own Mac the same way
10 days free, no card · Requires macOS 14 or later · Apple silicon