Smartwatch Blood Oxygen (SpO2) Accuracy Leaderboard
Reviewed by Sameer Sontakey, Co-founder and Chief AI & Algorithms Officer at Sensor Bio. ChatPPG is independent and device-neutral; it is not a Sensor Bio product.
In 36 people with COPD at rest, Garmin Vivosmart 4 (RMSE 3.3%) and Apple Watch Series 7 (RMSE 3.5%) had similar error versus a pulse oximeter. Both overestimated SpO2, and the authors advised against using them during pulmonary rehabilitation.
Which wearable is most accurate for SpO2?
23 published results from 14 studies in the wearable accuracy database. Each ranking below compares devices only on the same statistic against the same kind of reference. Results from different statistics or references are never ranked together.
Evidence by brand
- Apple Watch8 studies, n = 1243
- OuraNo published validation found
- WHOOPNo published validation found
- Garmin5 studies, n = 143
- Fitbit1 study, n = 30
- Pixel WatchNo published validation found
- Samsung3 studies, n = 93
- PolarNo published validation found
- Withings3 studies, n = 267
- AmazfitNo published validation found
- UltrahumanNo published validation found
- COROSNo published validation found
Blood oxygen (SpO2): root-mean-square error (% points)
Reference: Medical-grade pulse oximeter. Lower is better. 2 studies; a device tested in more than one study shows its median.
- 1Apple Watch Series 73.2 % points2 studies, n = 85 (3.5% (rest); 2.9%) · Støve et al. (2023); Jiang et al. (2023)
- 2Garmin Vivosmart 43.3 % points1 study, n = 36 (3.3% (rest)) · Støve et al. (2023)
- 3Garmin Venu 2S6.7 % points1 study, n = 49 (6.7%) · Jiang et al. (2023)
Blood oxygen (SpO2): root-mean-square error (% points)
Reference: Arterial blood gas (SaO2). Lower is better. 2 studies; a device tested in more than one study shows its median.
- 1Samsung Galaxy Watch41.8 % points1 study, n = 24 (RMSD 1.8%) · Browne et al. (2025)
- 2Apple Watch Series 75.82 % points1 study (Arms 5.82% (SaO2 < 88%)) · Jiang et al. (2026)
Blood oxygen (SpO2): mean absolute error (% points)
Reference: Medical-grade pulse oximeter. Lower is better. 1 study.
- 1Apple Watch Series 72.2 % points1 study, n = 49 (2.2% (RMSE 2.9%)) · Jiang et al. (2023)
- 2Garmin Venu 2S5.8 % points1 study, n = 49 (5.8% (RMSE 6.7%)) · Jiang et al. (2023)
Head-to-head results within single studies
Devices tested side by side in one study, on the same people and protocol.
Ranked: lower is better
SpO2 root mean squared error at rest vs pulse oximeter (COPD), lower is better
Støve MP, Graversen AH, Sørensen J (2023), Respiratory Care. doi:10.4187/respcare.10760
36 people with COPD · Reference: Contec CMS50D pulse oximeter · At rest, after 30-s sit-to-stand and after 6-min walk test
- 1Garmin Vivosmart 4Brand guide6.1% after sit-to-stand3.3%
- 2Apple Watch Series 7Brand guide4.1% after sit-to-stand3.5%
Reported result (not ranked)
Apple Watch 6 SpO2 agreement (single device)
Spaccarotella C, et al. (2022), Journal of Clinical Medicine. doi:10.3390/jcm11061467
265 screened (healthy, lung and cardiovascular disease) · Reference: Standard commercial pulse oximeter · Spot SpO2 readings, Bland-Altman analysis
- •Apple Watch Series 6Brand guider = 0.89Bias −0.23; LoA −3.49 to 3.04
All published results
Every SpO2 result in the database, including statistics that cannot be ranked. Download the full table as CSV for all rows. From the wearable accuracy database (v1.0.0).
Blood oxygen (SpO2)14 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| Apple Watch Series 7 | 3.5% (rest)Root-mean-square error | Medical-grade pulse oximeter | n = 36, Rest, COPD outpatients | Støve et al. (2023)Funding not reported |
| Garmin Vivosmart 4 | 3.3% (rest)Root-mean-square error | Medical-grade pulse oximeter | n = 36, Rest, COPD outpatients | Støve et al. (2023)Funding not reported |
| Apple Watch Series 6 | r = 0.89Pearson correlation | Medical-grade pulse oximeter | n = 265, Spot readings | Spaccarotella et al. (2022)Independent |
| Apple Watch Series 6 | −0.23 (LoA −3.49 to 3.04)Mean bias (signed) | Medical-grade pulse oximeter | n = 265, Spot readings | Spaccarotella et al. (2022)Independent |
| Apple Watch Series 7 | 2.2% (RMSE 2.9%)Mean absolute error | Medical-grade pulse oximeter | n = 49, Spot readings, with and without chronic pulmonary disease | Jiang et al. (2023)Independent |
| Apple Watch Series 7 | 2.9%Root-mean-square error | Medical-grade pulse oximeter | n = 49, Spot readings, with and without chronic pulmonary disease | Jiang et al. (2023)Independent |
| Garmin Venu 2S | 5.8% (RMSE 6.7%)Mean absolute error | Medical-grade pulse oximeter | n = 49, Spot readings, with and without chronic pulmonary disease | Jiang et al. (2023)Independent |
| Garmin Venu 2S | 6.7%Root-mean-square error | Medical-grade pulse oximeter | n = 49, Spot readings, with and without chronic pulmonary disease | Jiang et al. (2023)Independent |
| Withings ScanWatch | Highest missingness: 69.5% of attempts successfulStudy finding (no single number) | Medical-grade pulse oximeter | n = 49, Spot readings, with and without chronic pulmonary disease | Jiang et al. (2023)Independent |
| Apple Watch Series 8 | Met ISO RMS ≤ 4% at normoxia and SpO2 < 90%; lowest bias and RMSD of the threeStudy finding (no single number) | Medical-grade pulse oximeter | n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61 | Walzel et al. (2023)Independent |
| Samsung Galaxy Watch5 | Met ISO RMS ≤ 4% at normoxia and SpO2 < 90%Study finding (no single number) | Medical-grade pulse oximeter | n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61 | Walzel et al. (2023)Independent |
| Withings ScanWatch | Met ISO RMS ≤ 4% at normoxia and SpO2 < 90%Study finding (no single number) | Medical-grade pulse oximeter | n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61 | Walzel et al. (2023)Independent |
- Apple Watch Series 73.5% (rest)
Root-mean-square error · vs Medical-grade pulse oximeter · n = 36, Rest, COPD outpatients
Støve et al. (2023) Funding not reported
- Garmin Vivosmart 43.3% (rest)
Root-mean-square error · vs Medical-grade pulse oximeter · n = 36, Rest, COPD outpatients
Støve et al. (2023) Funding not reported
- Apple Watch Series 6r = 0.89
Pearson correlation · vs Medical-grade pulse oximeter · n = 265, Spot readings
Spaccarotella et al. (2022) Independent
- Apple Watch Series 6−0.23 (LoA −3.49 to 3.04)
Mean bias (signed) · vs Medical-grade pulse oximeter · n = 265, Spot readings
Spaccarotella et al. (2022) Independent
- Apple Watch Series 72.2% (RMSE 2.9%)
Mean absolute error · vs Medical-grade pulse oximeter · n = 49, Spot readings, with and without chronic pulmonary disease
Jiang et al. (2023) Independent
Root-mean-square error · vs Medical-grade pulse oximeter · n = 49, Spot readings, with and without chronic pulmonary disease
Jiang et al. (2023) Independent
- Garmin Venu 2S5.8% (RMSE 6.7%)
Mean absolute error · vs Medical-grade pulse oximeter · n = 49, Spot readings, with and without chronic pulmonary disease
Jiang et al. (2023) Independent
- Garmin Venu 2S6.7%
Root-mean-square error · vs Medical-grade pulse oximeter · n = 49, Spot readings, with and without chronic pulmonary disease
Jiang et al. (2023) Independent
- Withings ScanWatchHighest missingness: 69.5% of attempts successful
Study finding (no single number) · vs Medical-grade pulse oximeter · n = 49, Spot readings, with and without chronic pulmonary disease
Jiang et al. (2023) Independent
- Apple Watch Series 8Met ISO RMS ≤ 4% at normoxia and SpO2 < 90%; lowest bias and RMSD of the three
Study finding (no single number) · vs Medical-grade pulse oximeter · n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61
Walzel et al. (2023) Independent
- Samsung Galaxy Watch5Met ISO RMS ≤ 4% at normoxia and SpO2 < 90%
Study finding (no single number) · vs Medical-grade pulse oximeter · n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61
Walzel et al. (2023) Independent
- Withings ScanWatchMet ISO RMS ≤ 4% at normoxia and SpO2 < 90%
Study finding (no single number) · vs Medical-grade pulse oximeter · n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61
Walzel et al. (2023) Independent
11 more results in the database.
Read before comparing
- Pätz et al. (2022, Pediatric Cardiology, DOI 10.1007/s00246-022-02987-w) found Apple Watch 6 SpO2 not yet up to pulse-oximetry standard in 508 patients with congenital heart disease.
- Devices were less accurate at lower saturations, which is where SpO2 matters most clinically.
- Rankings combine studies only when they report the same statistic against the same kind of reference, and still differ in protocol and population. Treat close values as ties.
Download the wearable accuracy table (CSV)
Every leaderboard row with study, DOI, reference standard and sample size. We'll email you the link.
FAQ
Is Apple Watch blood oxygen accurate?
It depends on the population. Spaccarotella et al. (2022) found strong correlation (r = 0.89) with a pulse oximeter, but Støve et al. (2023) reported a 3.5% RMSE and overestimation in COPD, and Pätz et al. (2022) found it not yet at medical standard in congenital heart disease.
Is Garmin Pulse Ox more accurate than Apple Watch?
In the one head-to-head study (Støve et al., 2023), resting RMSE was similar (3.3% Garmin vs 3.5% Apple), but Garmin error grew more after exercise (6.1% vs 4.1%).