Apple Watch Ultra 3

Being measured is the interesting thing about this watch

Almost nothing on this site has been measured by anybody independent. Apple is the exception, and that alone puts it in an unusual position: its weaknesses are documented, while its rivals' weaknesses are merely unknown.

We want to be scrupulous about this. Every study below tested an earlier model, none tested an Ultra 3, and Apple revises its algorithms between generations. The results are the best published evidence about this platform and they are not measurements of this watch.

The sleep result

In the Brigham and Women's polysomnography study (n = 35, single-night inpatient PSG), the Apple Watch Series 8 achieved 75.0% four-stage agreement — respectable — but deep-sleep sensitivity of only 50.5%, the weakest of the three devices tested. It overestimated light sleep by 45 minutes a night and underestimated deep sleep by 43.

Fifty percent sensitivity on a two-outcome question is worth sitting with for a second. It means that on the specific stage most people care about, the watch and the clinical reference agreed about half the time.

The VO2 max result, which is newer and clearer

Two peer-reviewed validations of Apple's cardio-fitness estimate now exist, and they agree with each other.

Lambe et al. (2025) tested 28 participants across Series 5-9, SE 2 and Ultra 2 against indirect calorimetry. The watches underestimated VO2 max by 6.07 mL/kg/minon average (MAPE 13.31%), with limits of agreement from −6.11 to +18.26. The authors' conclusion is that the estimates "are not sufficiently accurate to inform clinical decision-making" while remaining useful for general fitness monitoring.

Caserman et al. (2024) found the same direction on a Series 7: lab 45.88 against watch 41.37 mL/kg/min, MAPE 15.79%, and an intraclass correlation of 0.47 — poor reliability. Their stratification is the useful part: the error grew with fitness, from 10.7% MAPE in the poor-fitness group to 21.5% in the excellent-fitness group.

If you are a trained runner buying an Ultra to watch your VO2 max trend, that last sentence is the one that matters. The published evidence says the estimate gets less accurate the fitter you are, and both studies note the sample skewed toward fit participants. The full VO2 max evidence page lays out what the number is actually good for.

What it is genuinely excellent at

Positioning hardware and display. Dual-frequency GNSS puts it in the top tier of consumer receivers, and no watch on this site has a better screen. As a general smartwatch it is not close — nothing here competes on notifications, apps, cellular or ecosystem.

Its weakness against dedicated sports watches is endurance: it scores 4 on our rubric where the Instinct 3 Solar scores 10. For a marathon that is irrelevant. For a hundred-miler it is disqualifying.

What we could not verify

• Anything about the Ultra 3 specifically.Every result above tested an earlier model. We are describing a platform's published record, not this watch.
• Heart rate during exercise. No Apple device appears in the 2025 placement study, so the direct wrist-optical comparison that covers Garmin and Whoop does not cover Apple.
• Positioning accuracy. The 2020 eight-watch study included an Apple Watch Series 4 — single-band, six years ago. It says nothing about a dual-frequency Ultra.
• Whether the sleep algorithm improved.Apple has revised it since the Series 8. Nobody has published a re-measurement, so "probably better" is a reasonable guess and not evidence.

Who should not buy it

Anyone buying it to fix their sleep data. The one published measurement says that is the thing it was worst at, and no firmware update rewrites a published measurement — the sleep accuracy ranking has better answers. Also anyone racing beyond about a day, where the battery is the binding constraint rather than any feature.