We Compared 6 Oven Thermometers Against a Calibrated Probe

Tested By LAB Nathan

Reviewed By HACCP Nouhayla A.

Protocol Updated: Jul 28, 2026

We put six oven thermometers in one oven, next to a reference probe accurate to ±[X]°F. At 350°F they read from [LOW]°F to [HIGH]°F. [N] of them we cannot separate, because the gap between them is smaller than our own resolution. And one accurate model turned out to be unreadable through a closed door.

We earn nothing from this page. All six thermometers were bought at retail price, and there are no affiliate links anywhere in this article.

Oven thermometers are accurate enough to be useful, but they are not equally accurate. At 350°F our six units read between [LOW]°F and [HIGH]°F. Our reference was a thermocouple probe with a manufacturer’s stated accuracy of ±[X]°F. Stated accuracy is the error the maker claims. Resolution is the smallest gap a test can honestly report, and ours is [T-RES]°F.

Below that number we group units together instead of ranking them. [N] of the six sit in one group. Most of these instruments are bimetallic dials: two bonded metals that bend as they heat, which moves a needle. They react slowly and drift after knocks. A second factor changed our order entirely, and that factor is readability through a closed door. You can also check the thermometer you already own at home.

• Are oven thermometers accurate enough to trust?
• How we made sure our reference was right
• What the six read at 325, 350 and 375°F
• Can you actually read it through the door?
• How to check the thermometer you already own

Are Oven Thermometers Accurate Enough to Trust?

Most oven thermometers sold today are bimetallic dials. Two bonded metals expand at different rates, the strip bends, and the needle moves. The design is mechanical and slow. A dial needs several minutes to settle after the surrounding air changes. That delay is called response time. A dial also shows hysteresis: the needle lands in a different spot coming down from a high temperature than climbing up.

Knocks and heating cycles shift it further. None of that makes a dial worthless. A dial reports a smoothed average, not the instant temperature. Independent laboratories reached the same conclusion before us. America’s Test Kitchen documented these limits of dial instruments, and their oven survey shows how far a cavity sits from its set point. Those are their samples and their dates, not our results.

Every reading in the oven data came from the reference probe we use here, so this article is the audit of our own instrument. Measuring your oven rather than your thermometer? That is the 20-minute test.

How We Made Sure Our Reference Was Right

Our reference is a [REF-MODEL] thermocouple probe with a stated accuracy of ±[X]°F. It [ships / does not ship] with a calibration certificate. We did not take that number on trust. We checked the probe in an ice bath, which holds 32°F. We checked it again at the boiling point of water, corrected for our altitude of [ALT] feet. Water boils at [BOIL]°F here, not 212°F.

Colorado State University Extension publishes that elevation correction, and skipping it invents an error that does not exist. From the stated accuracy we then set our resolution. Below [T-RES]°F we do not separate two instruments. We report them as one group instead. That threshold is why this comparison has fewer ranks than most. A comparison that hides its own resolution cannot be checked by anyone.

The protocol was simple. All six units went into one cavity at once, on one rack, with the reference probe among them. Ovens cycle on and off around the set point, so a single glance means little. ThermoWorks documents that cycling behavior, and their offset method averages the high and low of a stabilized cycle. We did the same. Positions were rotated between series, so no unit kept a favorable rack position. The Thermometer Check Card collects these checks.

What the Six Read at 325, 350 and 375°F

Each figure below is the deviation from the reference probe at that set point, in Fahrenheit. Swing is the spread between the highest and lowest reading inside one cycle. Response is the settling time. Units we cannot separate share a group letter.

Read the table as a spread, not a leaderboard. At 350°F the six units covered [SPREAD]°F from lowest to highest. That range matters more than any single number, because a baker works inside it without knowing. [MODEL-x] and [MODEL-y] differ by less than [T-RES]°F, so we report them together and refuse to call one better.

One unit read [Z]°F below the reference at all three set points, which puts it outside our useful range for baking. Swing tells a second story. A dial with a wide swing is not wrong. It is slow, and it smooths the cycle you wanted to observe. Two units needed more than [R] minutes to settle. That is long enough to mislead anyone who opens the door, looks, then closes it again.

Can You Actually Read It Through the Door?

This is the criterion nobody publishes, and it changed our conclusion. We photographed every unit from the real reading position: door closed, oven hot, camera at [DIST] feet and [ANGLE] degrees, identical for all six. Then we scored four things. Dial size and contrast. Parallax, which is the reading error you get when you view a needle from an angle. Condensation on the glass.

And orientation, because a dial that hangs sideways cannot be read at all. The result is blunt. Stated accuracy is not usable accuracy. [MODEL-a] sits in our tightest accuracy group and scored [SCORE] on readability. In practice, you have to open the door to use it. Opening the door drops the cavity temperature at the exact moment you are trying to measure it.

Here is the part that surprised me. I had a favorite before we started, the dial I have used for two years. On the table it does fine. Then I tried to read it through the closed door, and I could not. I had always opened the oven to look. The instrument was right. My use of it was not. Reading habits follow you onto a brand-new range.

Get the Thermometer Check Card. Free PDF for checking the thermometer you already own: ice bath, altitude-corrected boiling point, a log grid, and the resolution threshold that applies at home. Get the Thermometer Check Card.

How to Check the Thermometer You Already Own

You do not need to buy anything to find out whether your thermometer tells the truth. We ran this procedure on all six units before publishing it. It takes about twenty minutes.

1. Ice bath. Fill a glass with crushed ice, top with cold water, stir one minute. Submerge the sensing part without touching the sides. Wait until the needle stops. It should read 32°F.
2. Boiling point, altitude corrected. Bring water to a rolling boil. Look up the boiling point for your elevation in the Colorado State University Extension guide above. At 5,000 feet it is roughly 203°F, not 212°F. Reading against 212°F at altitude invents an error of nearly ten degrees.
3. Cavity comparison. Put your thermometer and a second one on one rack. Set the oven to 350°F and wait thirty minutes. Read both without opening the door.
4. Interpret. Note the offset at each checkpoint. A consistent offset is a correction you can apply mentally. An offset that changes direction between checkpoints is a unit you cannot correct.

One caution. A kitchen check is coarser than a bench check, so use a wider threshold. At home, treat anything under 5°F as unresolvable. A fan also changes what a dial sees, which is covered in reading a cavity in convection.

So which one would we keep? This is the one we would keep: [MODEL-k]. Here is the reading that decided it. It sits in the tightest accuracy group, and it is the only unit in that group we could read through a closed door at [DIST] feet. Two limits remain. We tested one sample per model, so unit-to-unit variation is unmeasured. Drift over time is not measured here, the most serious gap. We rerun the same six units in November 2026.

What to Remember

• Our resolution is [T-RES]°F. Below that gap we group instruments instead of ranking them.
• At 350°F the six units read from [LOW]°F to [HIGH]°F against a probe stated at ±[X]°F.
• Readability through a closed door changed the order. [MODEL-a] is accurate but hard to read.
• You can verify your own thermometer in twenty minutes, with ice water and altitude-corrected boiling water.

Next step: get the Thermometer Check Card and check the instrument already in your oven. Then see what home ovens measured, run the 20-minute test, or submit your oven.

Sources and Methodology

Protocol: six units, one cavity, three set points, rotated positions, rolling averages, declared resolution threshold. Reviewed by Nouhayla Azoumag, HACCP-certified, credentials on our methodology page. Photos by Nate, ovenlytic Test Kitchen. Drafted with AI assistance and reviewed before publication.

• America’s Test Kitchen, Finding a Good Oven Thermometer, updated 11 June 2025.
• America’s Test Kitchen, oven thermometer survey, 19 July 2026.
• ThermoWorks, Thermal Secrets of Oven Calibration, 19 July 2026.
• ThermoWorks, Checking Your Oven Temperature, 19 July 2026.
• Colorado State University Extension, High Elevation Food Preparation Guide, 19 July 2026.