Sous-vide circulators: what actually matters

Four things decide whether a circulator is worth buying, and none of them is what the spec sheet leads with. Physical buttons that outlive the app. Measured accuracy rather than the marketing's tenth of a degree. A clamp that actually fits your pot. And whether the machine survives a cook that runs past a workday. One of those stopped being theoretical in March 2026, when the software behind the original Joule shut down and its cooker, no dial, no display, no buttons at all, could no longer be told what to do (E·10, E·11).

So this guide walks the decision in the order you would reason it: heating power and what it actually buys, accuracy claims against independent measurements, app dependence, clamp fit, and whether the thing survives a three-day brisket. We read the manuals and spec pages ourselves and leaned on published lab measurements for the rest. Every number carries a marker like (E·01) pointing to the evidence ledger at the end.

The decision map

• If you cook mostly steaks and chicken in a stockpot, a few hours at a time, almost any mainstream circulator with on-device controls will do. Heat-up speed and app features barely matter.
• If you start from cold water in a container of ten quarts or more and don't want to wait, wattage class and circulation start to matter. Look at measured heat-up times, not the watt number alone (E·01).
• If you want the device to outlive its app, require physical on-device controls. The original Joule's app shuts down in March 2026 (E·10) and Anova app accounts created after August 21, 2024 pay a subscription (E·08).
• If you plan cooks of 24 to 72 hours, the deciding specs are a low-water shutoff, a lidded container and steam-tolerant construction, not accuracy claims (E·13, E·14).
• If your containers are thick-walled coolers or cambros, check clamp opening width and minimum water depth before anything else (E·06, E·12).

How much wattage you actually need

The watt number answers one question: how long you wait for tap-cold water to reach cooking temperature. It says nothing about maximum temperature or holding ability, and within a power class it is a poor proxy even for speed. In an independent lab test heating a 9.5 quart bath to 147°F, the 1100 W Joule Turbo took 32 minutes, the 1100 W Anova Precision Cooker 3.0 took 36, a 1000 W Inkbird tied a 1100 W Greater Goods at 39, and the 850 W Anova Nano needed 47 (E·01). Two units with identical wattage differed by four minutes. A lower-wattage unit matched a higher-wattage one. Circulation design and heater geometry are doing work the spec sheet doesn't show.

Widen the power gap and the correlation comes back. In a separate independent test to 132°F, a 1200 W Breville PolyScience HydroPro heated the bath in 15 minutes where the older Anova Precision Cooker needed 34 and the 1200 W Anova Pro 22 (E·02).

Measured data

Data: independent lab testing, cold tap water, uncovered container (E·01).

Every unit in that table will hold 147°F indefinitely once it gets there. Wattage doesn't buy better cooking. It buys minutes at the start, and starting from hot tap water erases most of the gap for free. An 800 W unit in an 8 quart pot is not a compromise. An 800 W unit asked to heat a 20 quart cooler from cold is.

If you batch-cook or work in large containers, stay in the 1100 W and up class and check measured heat-up, not rated. Weeknight cooks in normal stockpots can ignore the spec entirely and choose on controls and clamp instead.

Our head-to-head of the 1000 W Inkbird ISV-100W against the Anova Precision Cooker 3.0 walks that exact tradeoff on the same lab bench.

Which accuracy claims survive measurement

Manufacturers and reviewers use four different words as if they were one property, and buying decisions go wrong when they get mixed up.

• Precision is how finely the unit can be set and displayed, usually 0.1° increments. Nearly universal, and the cheapest property to print on a box.
• Accuracy is how close the actual bath temperature is to the number you set. This is where units differ.
• Stability is how little the bath wanders over time once at setpoint. A unit can be stable and consistently wrong at the same time.
• Uniformity is how much temperature varies across locations in the bath. That is circulation, not electronics. A strong pump matters more than a good sensor.

Anova's spec sheet declares "±0.1°C (0.2°F) accuracy" for the Precision Cooker (E·03). Real baths, independently measured, tell a messier story. The Joule Turbo ran 0.45°F cooler than setpoint on average, while an Instant Pot Accu Slim V2 averaged about 0.19°F over (E·04). An earlier Accu Slim unit measured 0.27°F under a 135°F setpoint but 0.72°F under at 147°F, so the offset itself changed with the set temperature (E·05). The Anova Nano, one of the cheapest units in its lineup, held an average discrepancy under a tenth of a degree once at temperature (E·06). At the bottom of the market the misses stop being fractional. A Gourmia held water about 2°F above its setpoint, and a VacMaster SV1 never reached setpoint at all, plateauing at 129°F against 132°F (E·07).

Measured data

Read those results with the four properties separated. Almost every unit tested was stable and precise. The differences were in accuracy, a fixed or temperature-dependent offset, and for the weakest units, in ever reaching setpoint. Measurements taken at six points in a bath (E·05) also fold uniformity into the number. A unit with weak circulation can look inaccurate at the edges of a large container while its sensor is fine.

For chicken breasts and ribeyes, a consistent offset of half a degree is irrelevant. It matters at edge-of-texture targets like 129°F steaks or 145°F custards, where a unit sitting 2°F hot silently moves you into a different result. If you cook at sensitive setpoints, favor units with measured sub-half-degree accuracy and strong circulation, then verify your specific unit once with a separate calibrated thermometer and apply the offset. That beats trusting any spec sheet, including the accurately-worded ones. Everyone else can treat the ±0.1° marketing war as noise.

What happens when the app goes away

The Joule shutdown from the top of this article is one half of the story. Anova is the other half: it kept the hardware independent and put the app behind a subscription instead.

Anova sells an optional Premium app plan, and in August 2024 it said accounts created before August 21, 2024 would stay free (E·08). Annoying, but survivable. Update, 28 October 2026: the fee amounts were removed from this page; the subscription guide linked below tracks what the plan covers today. Anova states its devices work without any subscription, and every model except the Mini has manual on-device controls (E·09). We checked the manuals; it's true. Our Anova subscription guide breaks that fee down model by model, including the one device in the lineup that cannot cook without the app at all.

The original Joule had no such fallback. No dial, no display (E·11). Owners who miss the migration to the Breville+ Cooking app, or Sage+ outside Breville's home regions, are left with hardware that cannot be told what to do (E·10). If that's you, our guide to the original Joule's app shutdown walks through the migration path, its one documented failure mode, and the choices left once you're on the other side of it.

If skipping the app question entirely appeals to you, we reviewed one dial-and-display circulator built around exactly that, the Greater Goods Kitchen Sous Vide (75/100, medium confidence). Our beginner's guide to sous vide without an app covers that whole no-app buying path, not just one model.

So the spec that matters is almost embarrassing: does it have buttons? If yes, app politics are an inconvenience. If no, the useful life of your hardware equals the manufacturer's willingness to keep servers running, and both Anova and Breville have now shown that willingness has limits and price tags. Anyone who keeps kitchen gear for a decade should treat on-device controls as non-negotiable and app features as disposable extras. If you genuinely want remote start and notifications, price in the Anova subscription for a new account.

Will it fit your pot

Attachment sounds trivial until it constrains everything else. Three mechanisms dominate.

• Adjustable screw clamps forgive pot walls of different thickness, at the cost of some fiddling. Independent reviewers dinged the Accu Slim's clamp screw for not rotating freely (E·05).
• Fixed-gap clamps are the strict ones. The Anova Nano's fixed twist clamp opens only 0.89 inches, which rules out thick-walled containers like coolers and some cambro-with-lid setups (E·06).
• Magnetic feet hold the Joule Turbo to "most magnetic steel pots and induction-ready cookware." Nonmagnetic or large containers need the separately sold Big Clamp accessory (E·12).

Add the water-level window. Anova's spec sheet lists a minimum-water shutoff sensor at the 65 mm level (E·03), which sets both how shallow your container can be and when the unit cuts out as water evaporates.

Decide what vessel you will actually cook in before choosing the circulator. Polycarbonate tubs and coolers, the natural homes of long cooks, are exactly where fixed narrow clamps and magnet-only bases fail, and where you may end up paying extra for an accessory the spec sheet never mentioned. If your kitchen runs on induction cookware, the magnetic base is a genuine convenience. For everyone else it is a constraint.

Will it survive the 48-hour cook

Short ribs, brisket and 72-hour experiments are why many people buy a circulator, and they are exactly the duty cycle where failures cluster. Owner reports show two distinct mechanisms, and neither is visible on a spec sheet.

Evaporation ends cooks early. In one forum report, a 24-hour cook at 180°F in an uncovered vessel lost most of its water while unattended. The circulator hit its low-water sensor and shut off mid-cook, leaving the food cooling for hours (E·13). The shutoff did its job. The failure was the uncovered container. Community consensus in the same thread: "for longer cooks you really want a covered vessel" (E·13). A low-water shutoff is a needed safety feature, not a substitute for a lid.

Steam kills electronics. Owners report repeat failures concentrated around long, hot cooks. Units die outright, and in the worst pattern they lose temperature control while displaying normal readings. One unit boiled its bath 40 hours into a 60-hour cook while showing 140°F on target (E·14). The suspected mechanism appears in the manufacturer's own support language quoted by owners: "steam may have already gotten into the head of your unit" (E·14). Manufacturers do rate some hardware for this duty. Anova markets its Pro model on 10,000 hours of continuous cooking (E·14). Those are declared characteristics, though, and reliability is exactly the claim a manufacturer cannot certify for itself.

For long cooks, the setup matters as much as the unit. A lidded or film-covered container, water topped well above the minimum line, and the circulator positioned so bath steam doesn't vent into its head. A unit that can boil the bath while displaying setpoint is also a food-safety argument for checking any multi-day cook with an independent thermometer. If cooks past 24 hours are your main use case, weight low-water shutoff behavior, sealed construction and duty-rated models above heat-up speed and app polish, and treat the container and lid as part of the purchase.

None of this closes every gap. Forum threads document vivid failure modes (E·13, E·14) but are self-selected, so no source here can say what fraction of units actually fail, or whether one brand outlasts another; the accuracy figures above (E·04 through E·07) each describe a single retail unit, and two Accu Slim versions already diverged in the same lab (E·04 vs E·05). Anova's free-forever promise (E·09) and the Breville+ app's role as the Joule migration path (E·10) are both current policy, not guarantees, and the Joule shutdown is the reason to read them that way. Every measured heat-up time here also comes from an open container (E·01, E·02); a lid will shrink the gaps between wattage classes by an amount nobody has published. What the ledger does establish, reliably, is which of these questions to ask before you check a price.

The short version

Buy on buttons first, measured accuracy second, clamp fit third, and long-cook setup fourth. The watt number on the box matters least. The app matters most on the day it disappears, and both major brands have already shown you that day exists.