The short version
- Yes, most do. 85% of the 13,336 certified mini split systems meet the ENERGY STAR cold-climate spec.
- The typical system keeps 83% of its heating output at 5 °F (−15 °C). The weakest 10% keep 68% or less; the best 10% keep 100% or more.
- About 19% of systems still give their full rated heat at 5 °F. If you live somewhere with real winters, filter for those.
Why heat pumps lose capacity in the cold
A heat pump doesn’t make heat, it moves it. In winter it pulls heat out of the outdoor air and pumps it inside. The colder the outside air, the less heat there is to grab and the harder the compressor has to work. Two things drop as temperatures fall:
- Capacity: how many BTU per hour the unit can deliver.
- Efficiency: how much heat you get per unit of electricity (the COP).
Manufacturers have to report both at fixed test temperatures: 47 °F, 17 °F and, for cold-climate certification, 5 °F. That makes it possible to compare every system on the same footing.
What the certified data shows
We took the rated heating output at 5 °F for every certified system and divided it by the output at 47 °F. That gives the share of heating power left on a very cold day.
| Measure | All certified systems |
|---|---|
| Systems analysed | 13,336 |
| Meet the cold-climate spec | 85% (11,401) |
| Median heat left at 5 °F | 83% |
| Weakest 10% keep | 68% or less |
| Best 10% keep | 100% or more |
| Full rated heat at 5 °F | 19% of systems |
| Median efficiency (COP) at 5 °F | 2.00 |
A COP of 2.00 means the typical system still delivers about two units of heat for every unit of electricity at 5 °F, roughly twice as efficient as electric baseboard heat even on a frigid day.
The best 12,000 BTU systems for cold weather
These 9,000 to 12,000 BTU systems keep the most heating capacity at 5 °F, according to their certified ratings:
| System (outdoor / indoor model) | BTU/h | SEER2 | HSPF2 | Left at 5 °F | Refrigerant | Score |
|---|---|---|---|---|---|---|
| Fujitsu H-SeriesAOU9RLFF / AGU9RLF | 9,000 | 26.0 | 11.8 | 149% | R-410A | 81 |
| Fujitsu H-SeriesAOU9RLFFH / AGU9RLF | 9,000 | 26.0 | 11.6 | 149% | R-410A | 81 |
| Gree FairyGWH09ACD-D3DNA1A/O / GWH09ACD-D3DN***/I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Gree LomoGWH09ACD-D3DNA1A/O / GWH09QD-D3DN***/I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Gree AmberGWH09ACD-D3DNA1A/O / GWH09YD-D3DN***/I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Gree BoraGWH09ACD-D3DNA1A/O / GWH09AAD-D3DN***/I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Gree MuseGWH09ACD-D3DNA1A/O / GWH09AFD-D3DN***/I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| ACD LOMO 38SEERUHD09KCH38SB-O / UHD09KCH38SB-I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Wabban LOMO PlusBB09HQ4D6DO / BB09HQ4D6DI | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Blueridge SaphireBMKH0938/O / BMKH0938/I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Ac Pro.Com LOMO PlusAWH09ACD-D3DNA1A/O / AWH09QD-D3DND6B/I | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
| Trane 38 Series4TXK3809B10NU** / 4MXW3809B10NU** | 9,000 | 27.0 | 11.0 | 144% | R-410A | 79 |
Some keep more than 100%. That isn’t a typo: some cold-climate units are designed to run their compressor faster in the cold, so they deliver more than their 47 °F rating.
How to use this when you shop
- Size for the cold, not the brochure. If your room needs 10,000 BTU/h of heat on the coldest night and a unit keeps 70% at 5 °F, a “12,000 BTU” model delivers only about 8,400 BTU/h when you need it most. Our sizing calculator shows the heat you need at 5 °F.
- Look for the 5 °F number, not just “works down to −13 °F”. Many units run at very low temperatures but deliver much less heat there.
- Filter the database: in the model database you can tick “cold-climate only” and sort by “% left at 5 °F”.
Methodology
Data: U.S. EPA ENERGY STAR certified central air conditioners and air-source heat pumps, product type “mini or multi split” (2026-11-17). Ratings follow the AHRI 210/240 test procedure. We excluded no systems; a few listings without a 5 °F rating are left out of the percentages. Full methodology.
