Somebody on Reddit recently suggested that the AirGradient Go and ONE might report CO₂ on different scales. If you have both monitors and they show different numbers, it is an understandable question. Are you seeing normal differences between individual sensors, or does one model respond differently as the CO₂ concentration changes?
Rather than leave it as a discussion about specifications, we ran a comparison with two Go monitors and two ONE monitors. We also wanted a noticeable change in CO₂, so the team helped by jumping around. More on that part below.
The results are excellent. Across 407 matched minutes, the Go average was 7.52 ppm higher than the ONE average, with a root mean squared difference of 11.84 ppm. The two groups followed the same broad rise and fall in CO₂, rather than behaving as though they were measuring on substantially different scales.
For me, this is the consistency we should expect from our products. Someone using a Go alongside a ONE should not have to wonder whether the two numbers mean different things. This test gives us a strong, practical reason to be confident in their agreement.
What's important to point out is that you might need to do a manual baseline calibration before you start this comparison to make sure they both start from the same absolute CO2 concentration.
What Did We Test?
We compared four monitors on 21 October 2026, using an export window from 12:50 to 20:00 Bangkok time. This four-monitor dataset contains 1,694 records:
- AirGradient Go, P-1PSG: units ending 8c90 and a248.
- AirGradient ONE, I-9PSL-DE: units ending 683c and 4e20.
For this comparison, we used the CO₂ (ppm) raw field from the export. We did not apply an additional correction to bring one model into line with the other.
We wanted to look beyond a single reading. A fixed difference between two monitors is one thing; a difference that grows as concentrations rise is another. Recording the rise and fall of CO₂ lets us examine both.
Why We Chose Optical NDIR for the Go
This comparison also connects to a bigger design decision. In our earlier article, CO₂ Sensor Technologies: Optical NDIR vs Photoacoustic vs Thermal Conductivity, we tested the Senseair S12, Sensirion SCD4x and Sensirion STCC4 during different real-world conditions.
The S12's conventional optical NDIR approach was the most stable and predictable option in those tests, particularly as conditions changed. That mattered to us because a portable monitor does not spend its whole life sitting on a shelf. It gets carried between environments and switched on and off.
Based on that testing, we believe optical NDIR is the best choice for the way we want the Go to work. The Go uses the Senseair S12, which measures CO₂ through infrared absorption. Not every portable monitor makes the same choice. Atmotube PRO 2, for example, uses the much cheaper Sensirion STCC4, a thermal-conductivity CO₂ sensor.
Our earlier tests explain why we prefer the optical approach for the Go. This four-monitor experiment in this article answers a different question: how closely our Go and ONE units agree.
First, a Look at the Four Monitors
If you wonder why you primarily see blue line, it is because all lines are basically on top of each other.
The first thing to notice is how closely the traces follow the same broad pattern. There are differences between individual units, but the Go monitors do not form one clearly separated group while the ONE monitors form another. Both groups follow the rise in CO₂ and the subsequent decline.
That is a very good result. We are not just seeing agreement at one quiet moment: the monitors follow a changing signal over the recording. To put numbers behind that visual impression, we also compared their readings directly.
How We Created the CO₂ Spike
The team provided the extra CO₂ by jumping around. This was not a new calibration method, although it did give everyone a short workout.
People exhale CO₂, and physical activity increases its generation. That gave the monitors a changing signal to follow, rather than leaving us with a long recording of nearly constant air.
How Close Were the Readings?
For a closer look, we compared the average of the two Go monitors with the average of the two ONE monitors. The scatter plot and statistics below use these same four units.
For each calendar minute, we took the latest valid reading from each monitor and averaged the readings within each model. We included a minute only when every monitor required for that comparison had a valid reading. We did not fill missing readings by interpolation.
| Metric | Two Go monitors versus two ONE monitors |
|---|---|
| Matched minutes | 407 |
| Mean difference, Go minus ONE | +7.52 ppm |
| Root mean squared difference | 11.84 ppm |
| Mean absolute difference | 10.51 ppm |
| R², squared Pearson correlation | 0.99867 |
| Fitted slope, Go average versus ONE average | 0.9806 |
| Fitted intercept | +20.11 ppm |
The Go average was 7.52 ppm higher over the matched minutes which is a small average offset for these units.
The root mean squared difference, or RMSE, describes the size of the paired differences while giving more weight to larger disagreements. It was 11.84 ppm. Against the much larger concentration changes during the test, I consider that a strong result. These are differences between model averages, not individual-device accuracy specifications.
The average of ONE units 683c and 4e20 is shown on the horizontal axis; the average of Go units 8c90 and a248 is on the vertical axis. Each point represents one matched minute. The dashed line shows equal readings.
The points sit close to the equal-reading line over a range of approximately 370–1,340 ppm for these model averages. The high R² also shows that the two averages follow the changing CO₂ concentration closely.
What Can We Take From This?
My takeaway is straightforward: this is an excellent result for the Go and ONE. We tested two units of each model, saw them follow the same broad CO₂ changes, and found only a small disagreement between their model averages. That is substantially more convincing than two displays happening to match in a photograph.
This does not mean two monitors will always show exactly the same number. If you notice a persistent difference, look at its size and pattern, measurement timing, placement and calibration settings. But for this test, the overall finding is clear: the two product families agreed very well.
We are happy to put those results out for people to examine. And if another useful question requires the team to jump around again, at least we now know that counts as product testing.
Our Support is Always there
In case you have different results when comparing the Go with the ONE, make sure you do a manual CO2 calibration. If you still see problems, please reach out to our support team, who are happy to help you.
