Methodology
How we rate, and what the labels mean
Read this before you trust a number on this site. The short version: every score is defensible from data we cite, the byline tells you whether a human touched the machine, and where we don't know something we say so instead of guessing. The Linux Readiness Score is computed, not felt, and the formula is below so you can argue with the arithmetic.
The seven scores
Seven categories, each 0 to 10: Linux readiness, battery, screen, build, keyboard and trackpad, performance, value. None of them are vibes. Each one has to point back at something you can check.
- Linux readiness: computed from the per-distro hardware table, not hand-set. Full formula below.
- Battery: real-world runtime where we have it, not the manufacturer's lab figure. "Lenovo says 18 hours, we logged 9 with WiFi on" is the kind of line that defends this number.
- Screen: panel type, measured or reported nits, refresh rate, and whether it uses PWM, which matters if your eyes are sensitive to it.
- Build: chassis material, hinge, flex, port selection, serviceability. A soldered SSD costs points and we say why.
- Keyboard and trackpad: travel, layout quirks, trackpad size and driver behaviour, the things you only notice after a week.
- Performance: the CPU and GPU class against what the machine is actually for, not a raw benchmark divorced from thermals on battery.
- Value: the score against the real price in your region, recomputed when the price moves. A great laptop at a bad price is a bad value score.
What defends each one is the cited source next to it: a measured figure, a community report, or a first-hand test with a date. If a number has no defence, it does not ship.
Where the data comes from
- Manufacturer specifications, labelled as claims, never as measured fact
- Community reports: ArchWiki, linux-hardware.org, distro forums, Reddit threads
- Real-world battery and thermal figures over manufacturer claims wherever we have them
- First-hand testing for the machines we own and ran ourselves
Two kinds of byline
Every model page carries one of two labels, and the difference is not cosmetic.
Reviewed by Dennis means the machine was run first-hand, with a test date on the page. First person is allowed there because it happened. This tier is small on purpose: the Razer Blade 16 (2024), the Redmibooks, a handful of Chinese laptops, and the distro tests done on them.
Compiled & edited by Dennis means the page was built from public sources and community data, then edited by a human. No hands-on claim, no invented first-person experience, no "I tested" for a machine nobody plugged in. Most of the catalogue sits here, and that is honest rather than a weakness. A spec sheet read carefully is still useful. It is just not the same thing as having owned the laptop, and we will not blur the two to look more expert than we are.
How fresh it is
Every model carries a "last verified" date. Pages are re-checked on a rolling quarterly cadence. If a field has not been verified, the page says "unverified" rather than filling the gap with a guess. A score is only as current as the date next to it, and laptop support, Linux support especially, moves under you.
Staged indexing, said plainly
New long-tail model pages can start un-indexed until they have enough depth and signal, then flip to indexable. This is staging, not cloaking. The same content is shown to every visitor and every crawler. The only thing that changes is whether we ask search engines to list a thin page before it is worth listing.
How we score Linux readiness
Every laptop here gets a Linux Readiness Score from 0 to 10, one decimal. It is not a feeling. It is computed the same way every time, from the same data you can see in the per-distro table further down each model page. Same data in, same number out. If you think the number is wrong, you can check our arithmetic, and you should.
Two things go in. First, the grade for a clean install on each distro we checked: works out of the box, needs a couple of known tweaks, or problematic. Second, six pieces of hardware that actually decide whether you will be happy: suspend, wifi, audio, webcam, bluetooth, fingerprint. Suspend and wifi carry the most weight, because a laptop you cannot close the lid on, or that cannot reach the internet, is not a working laptop. Fingerprint carries almost none. Most Linux-first machines ship without a reader on purpose, and password login always works, so a missing reader does not drag the score down.
| Component | Weight | Why that weight |
|---|---|---|
| suspend | 0.28 | The biggest real-world Linux failure. Broken s2idle kills the battery in a bag. Hardest fault for a user to work around. |
| wifi | 0.26 | No wifi is close to no laptop, and you cannot even download the fix. Edged out by suspend only because dongles and tethering exist. |
| audio | 0.16 | Speakers and mic. Blocks calls, but degrades to USB or Bluetooth audio. Serious, not fatal. |
| webcam | 0.14 | Genuinely blocking for many, but the most substitutable: a 20 euro USB cam fixes it. |
| bluetooth | 0.10 | Mouse, headphones, files. Annoying when dead, almost always a wired path. |
| fingerprint | 0.06 | A login convenience. Often absent by design on Linux-first machines. Lightest on purpose. |
The grade and the hardware split 45 / 55. The hardware components get the larger share because they are the falsifiable evidence; the grade is a human summary and can drift. The grade keeps a heavy 45% anyway, because it catches things the six boxes do not: GPU driver pain, thermal throttling on battery, an unusable trackpad. Neither term is allowed to fully override the other. A laptop where five of six boxes are green but the clean-install grade is problematic, the Dell XPS 13 with a dead webcam is the real case, does not float up to a 9. The 45% grade anchor is what stops that, and it is deliberate.
We score the machine at its best documented distro and we tell you which one. If a laptop is flawless on Fedora but nobody has tried it on Arch, that is a good Linux laptop, and averaging in a distro nobody tested would only punish it for the size of our test bench. The per-distro table tells you the rest.
Now the honest part. When we have not verified something, we mark it unknown, and unknown never helps the score. It is dropped from the math, not counted as a pass and not counted as a fail. A laptop nobody has tested is not a good Linux laptop. It is an untested one, and those are not the same thing. If we know too little to score a machine fairly, a known clean-install grade plus at least three of six known components is the bar, we print "Not rated" instead of inventing a number. You will see that on the MacBook Air M4: Linux on Apple Silicon means Asahi, Asahi barely boots on M4, so there is nothing honest to score yet.
Next to every score is a confidence level: high, medium, or low. That is a separate number on purpose. It tells you how much we actually know, not how good the machine is. A 6 we are sure of and a 6 we half-guessed are different, and you get to see which one you are looking at. NVIDIA laptops are capped at medium even when the data is complete, because the proprietary driver stack changes under you and a score from three months ago may not survive the next kernel.
The limits, stated plainly. The score is only as current as our last check date, shown on the page. Linux support moves; where an upstream tracker is the live truth, such as Asahi, we link it and you should trust that over us. We do not hide behind the formula either. The clean install grade is a human call and it is weighted heavily, so our judgement is still in the number, just out in the open where you can disagree with it. If you only need wifi and suspend and you do not care about the webcam, read the per-distro table, not just the headline. That table, not the single number, is the thing to act on.
Disagree with a number?
Good. Every score points at a source and the Linux formula is fully public, so a disagreement is an argument about evidence, not taste. Tell us where the arithmetic or the source is wrong via the contact page and we will fix it or show our working. See also the people and impressum, how we intend to make money (nothing is monetised yet), and the privacy policy.