How it works: where the numbers come from
Every number here is an official test rating. We gather them, group each model with others of its size and type, and explain them in plain words. Nothing is estimated for a particular home.
Covers heat pumps sold new today (R-454B and R-32 refrigerant). See older R-410A models · Updated
Three sources
AHRI Directory of Certified Product Performance
The official directory of certified ratings. Every model here is listed in it.
Last updated
NEEP Cold Climate Air Source Heat Pump List
A list of cold-climate heat pumps with test results at 5°F and colder.
Last updated
ENERGY STAR certified heat pumps
The U.S. EPA’s list of ENERGY STAR certified heat pumps, including the cold-climate label.
Last updated
Like for like
Same size, same type
Each model is only compared with models of the same size class (by rated cooling output) and the same type (ducted or ductless). We call them similar heat pumps.
- Up to 1.5 tonsunder 21,000 Btu/h
- 2 tons21,000–26,999 Btu/h
- 2.5–3 tons27,000–38,999 Btu/h
- 3.5–4.5 tons39,000–56,999 Btu/h
- 5 tons and up57,000 Btu/h and up
Top cold-climate picks
Keep the heat, then rank on three ratings
A pick must keep at least 70% of its heating output when it is 5°F outside, compared with a mild 47°F day. Units that keep less, or have no certified 5°F output, are left out.
The rest are ranked on three ratings at once, each against units of the same size and type: how much heat it keeps at 5°F, its efficiency at 5°F, and its winter-long efficiency (HSPF2). The score is the plain average of the three, out of 100.
Efficiency at 5°F alone can mislead: it is measured at whatever heat the unit puts out, so a unit that puts out little heat in the cold can still post a high number.
The words we use
For a model, within its size class
- Top 10% the top 10%
- Great the top quarter
- Good better than half
- Below the middle the lower half
For a brand: how many similar heat pumps its typical model beats
- Well above average60% or more
- Above average55% to 60%
- About average45% to 55%
- Below average40% to 45%
- Well below averageunder 40%
The ratings
COP at 5°F
Efficiency on a 5°F day
How much heat the unit makes from each unit of electricity when it is 5°F outside.
An electric baseboard heater has a COP of 1, so a COP of 2 means twice the heat from the same electricity in the 5°F test.
Capacity at 5°F vs 47°F
Heating power kept at 5°F
How much of its mild-weather heating power the unit still delivers at 5°F.
At 100% the unit puts out as much heat at 5°F as at 47°F in the test. Lower means less heating output in deep cold.
HSPF2
Winter-long heating efficiency
Heating efficiency over a whole winter, measured with a federal test.
Like miles per gallon for heating: higher means more heat per unit of electricity over the test season. The federal minimum for a new split heat pump is 7.5.
SEER2
Summer-long cooling efficiency
Cooling efficiency over a whole summer, measured with a federal test.
Like miles per gallon for cooling: higher means more cooling per unit of electricity over the test season. The federal minimum for a new split heat pump is 14.3.
Common questions
- Where do the numbers come from?
- From the AHRI Directory of Certified Product Performance (the official test ratings), NEEP’s cold-climate heat pump list and ENERGY STAR. The AHRI data was last updated Oct 9, 2026.
- What counts as one model?
- One outdoor unit. Most are certified with several indoor units; each model’s numbers come from one of those certified pairings, the one with the best efficiency at 5°F, never a mix.
- What about an outdoor unit sold with both ducted and ductless indoor units?
- A model sold with both ducted and ductless indoor units is rated in each type with its best pairing of that type. It is compared with central (ducted) systems through its best air-handler pairing and with mini-splits through its best wall or ceiling unit, and each list names the indoor unit it is ranked on. It still counts as one model. Multi-zone (“mixed”) systems are compared with mini-splits, but when you look up models, “Ducted (incl. mixed)” includes them, since they can serve ducted rooms too.
- Why does a pick say “also sold as”?
- The same unit is often sold under several brand names (Daikin, Amana and Goodman, for example). When the same manufacturer, or the same model number, is behind two listings with identical ratings, the lists show it once and name the other brands, so one unit never takes several places. Two different companies with equal ratings are never merged.
- Why only compare heat pumps of the same size and type?
- Smaller units and ductless mini-splits test higher than big central systems. The typical cooling rating (SEER2) is 16.0 for the smallest ducted units and 15.8 for the largest. Comparing across sizes would mostly compare sizes, so every list and score here stays within one size class.
- How are the top cold-climate picks chosen?
- Only units that keep at least 70% of their heat at 5°F, ranked on heat kept, efficiency at 5°F and seasonal efficiency (HSPF2), against units their size. Each of the three is the share of same-size, same-type models the unit beats, and the score is their plain average, so a unit has to do well on all three. A unit without a certified heat output at 5°F does not qualify.
- What is the all-around best?
- Balances heating (5°F output and efficiency, HSPF2) and cooling (SEER2, EER2), each rating against units their size. Where the cold-climate picks ask whether a unit holds up in deep cold, the all-around best asks whether it is good at everything: half heating, half cooling, each rating as the share of same-size, same-type models the unit beats. A unit needs all five ratings to be listed, and there is no minimum heat kept at 5°F; each pick shows whether it is also a cold-climate pick.
- Why not just rank by efficiency on a 5°F day?
- Because that number is measured at whatever heat the unit puts out at 5°F. A unit can be very efficient while putting out little heat when it is that cold, and then it needs backup heat sooner. So the picks first require a unit to keep most of its heat, and then also weigh how much heat it keeps and its winter-long efficiency.
- Is a top cold-climate pick the right one for my home?
- Not necessarily. These are certified test ratings, not a recommendation for your home. The right size comes from a load calculation by a contractor. Installation quality matters as much as the model.
- Why are some models never picked as "best"?
- Two reasons: its own cold-weather test numbers are inconsistent (a colder reading comes out more efficient than a warmer one, beyond the normal dip most heat pumps show partway down), or AHRI no longer lists it, so it cannot be bought. Either way it still shows its real numbers everywhere else; it just cannot win a "best" spot.
- What is a QMID?
- A QMID (Qualified Manufacturer Identification Number) is a short code the IRS gives a manufacturer that has registered as a qualified manufacturer. For heat pumps placed in service in 2025, the federal Energy Efficient Home Improvement Credit (25C), claimed on IRS Form 5695, needs equipment from a qualified manufacturer, identified by a product number that starts with its QMID. Congress ended that credit for equipment placed in service after December 31, 2025. Here a QMID on file is also what lets a brand be ranked against other brands. It says nothing about how well a model performs. Look up a manufacturer’s QMID
- Which brands get a score?
- Brands whose manufacturer has a QMID on file (the ID the federal 25C tax credit needs): 30 of 343, once at least 10 of their models have every rating. Every other model is still listed.
- What do good, great and top mean?
- They place a rating within its own size class: good is better than half of the class, great is the top quarter, top is the top 10%.
- Does a brand with more models sell more?
- No. The counts are certified models, not sales or installations.
These are certified test ratings, not a recommendation for your home. The right size comes from a load calculation by a contractor.
The advanced view has the exact ranking rules, every size class with its medians, data exclusions, sync counts and the full criteria of each standard.
Open the advanced view