2026 EU heat-pump refrigerant leakage risk by type: R290, R32 and the rest

What the 2026 EPREL refrigerant mix looks like

The current EU heat-pump catalog is vast: 60,989 listed models from 777 manufacturers, with an average SCOP of 4.55, average declared power of 9.3 kW and average outdoor sound level of 61.3 dB (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). The live EPREL catalog and our market index snapshot make clear that the refrigerant mix inside that universe is anything but balanced.

Using declared EPREL refrigerant codes, R32 models account for 13,935 listings, equal to 22.85% of all catalogued models (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). R410A models add 1,896 listings, or 3.11% of the total, while R290 models account for 537 listings, or 0.88% (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). Natural refrigerants as a group reach just 3.27% of the catalog (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API).

That leaves two important caveats. First, the declared refrigerant counts do not sum to the full market total, so EPREL refrigerant coding is incomplete in this snapshot (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). Second, the registry shows several variant spellings and obvious miscoding entries, including R290A, R290a, R410 and R41OA, each with small counts from 1 to 10 listings (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). For buyers checking actual stock rather than taxonomy, the filtered refrigerant catalog is more useful than any single raw code count.

A simple ranking of the declared refrigerants shows how concentrated the market still is:

\*R420A, R423A, R23, R421A, R411A, R35, R422A, R41OA, R413A, R425A, R419A, R33, R417A, R332 each appear once except the combined single-listing total shown here (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API).

The practical reading is straightforward: the largest leakage-risk exposure is not where the best refrigerants are heading, but where the listing volume already sits. Today that is still overwhelmingly R32.

Leakage-risk proxies: GWP, flammability and phase-out timing

This article uses a leakage-risk proxy rather than measured field leakage rates, because the registry does not record real-world refrigerant loss events. What it does record, and what the corpus adds, are refrigerant GWP values, flammability classes and phase-out timing from the current EU F-gas framework (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes).

At the low end of GWP, R290 and R600a are both listed at GWP 0, while R717 is also at 0 and R744, R1234yf and R1234ze(E) are at GWP 1 (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes). At the high end, R410A has GWP 1,924, the highest figure in the reference table, followed by R407C at 1,624 and R134a at 1,300 (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes).

Among refrigerants that actually show declared EPREL usage in this snapshot, the highest-GWP high-volume code is R410A at 1,896 listings with GWP 1,924, while the lowest-GWP high-volume code is R290 at 537 listings with GWP 0 (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes; market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). R32 sits between them at GWP 771 with 13,935 declared listings (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes; market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API).

Flammability complicates any simplistic reading. R290 is A3, meaning highly flammable; R32 is A2L, lower flammability; R410A and R134a are A1, non-flammable in this classification scheme (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes). So a lower climate impact proxy does not automatically mean lower handling complexity.

Phase-out timing adds another layer. The corpus lists phase-out dates of 2025-01-01 for R407C and R410A, 2026-01-01 for R134a, and 2027-01-01 for R32 (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes). For regulatory context, readers can cross-check the current EU F-gas regulation and the public EPREL portal.

The resulting proxy is useful but bounded: high GWP plus high installed listing volume implies greater market exposure if leaks occur; high flammability implies stricter installation and servicing controls; and short phase-out timing implies weaker forward compatibility. On that basis, R410A and especially R32 carry much of today’s exposure simply because they are common, while R290 scores far better on GWP but demands more attention to A3 installation practice.

Which heat-pump types pair best efficiency with lower-risk refrigerants

By type, the EPREL market is led by air-water heat pumps at 30,452 models, followed by air-air at 21,065, hot-water heat pumps at 9,228, ground-water at 213 and water-water at 31 (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API).

Efficiency and acoustic averages differ sharply:

(type_efficiency / EPREL Public API · type aggregation)

On pure efficiency, water-water is the standout with average SCOP 6.15 and the lowest average outdoor noise at 42.0 dB, but it is also tiny at 31 models, so it does not define market-wide exposure (type_efficiency / EPREL Public API · type aggregation). Ground-water is next on average SCOP at 4.77 across 213 models (type_efficiency / EPREL Public API · type aggregation). Air-water sits lower at 4.54, but its scale makes it the decisive segment for any refrigerant transition (type_efficiency / EPREL Public API · type aggregation).

The corpus does not provide a cross-tab of type by refrigerant across the full catalog, so it cannot numerically identify which type contains the highest share of low-GWP refrigerants. What can be said is narrower: because air-water is the largest type at 30,452 models and R290 is a low-GWP refrigerant with explicit catalog filtering, the most commercially relevant efficiency-versus-leakage-risk contest is inside the air-water segment, not in niche water-water stock.

R290 vs R32 in air-water: model counts and SCOP at the top end

This is the comparison most buyers now want, and the corpus only answers it partly.

Across the whole market, R32 has 13,935 declared listings while R290 has 537 (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). Within air-water specifically, the provided corpus does not include average SCOP by refrigerant, nor a refrigerant-by-type listing count. So the registry snapshot here cannot support a numeric statement such as “R290 air-water averages X while R32 air-water averages Y”. The article should not pretend otherwise.

What the corpus does show is that the top SCOP end of the market is dominated by water-water and air-water specialists rather than by any explicitly identified refrigerant code, because the top-model records in this extract do not carry refrigerant values; they are null for all 15 entries (top_models / EPREL Public API via Househeating Pulse catalog). That means the registry extract cannot count how many R290, R32 or other refrigerants appear in the top 15 overall or top 15 air-water SCOP ranks. The dedicated probes for top air-water R290 and top air-water R32 return empty arrays in this corpus, which again means no ranked numeric comparison is available here (top_models / EPREL Public API via Househeating Pulse catalog).

Even so, the top of the SCOP table is worth locating. The overall leader is Risch Kälte- und Klimatechnik GmbH OH I 4esr TWW W/W at SCOP 7.0 (top_models / EPREL Public API via Househeating Pulse catalog). Other front-runners include Hoval Aktiengesellschaft 42 -Thermalia® twin (26) GW at 6.97, Waterkotte GmbH CTC EcoTouch 525 (water/water) at 6.97, Waterkotte GmbH EcoTouch DS 5034.5 T (water/water) at 6.97, and Master Therm tepelná čerpadla s.r.o. AQ30I-0WW at 6.97 (top_models / EPREL Public API via Househeating Pulse catalog). The broader top SCOP leaderboard remains the best route for fresh model-level checks.

So the strongest evidence available here is not “R290 beats R32 by X SCOP points in air-water”, because the corpus does not contain that figure. It is instead that low-GWP refrigerants are still underrepresented in declared listing volume, while the efficiency frontier is populated by premium specialist units whose refrigerant fields are missing in this extract.

Where legacy refrigerants still dominate the listings

The exposure story is much clearer than the top-end performance story.

R32 alone has 13,935 declared listings, over seven times the 1,896 listings of R410A and roughly twenty-six times the 537 listings of R290 (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). Put differently, the biggest declared refrigerant pool in EPREL is tied to a refrigerant with GWP 771 and a listed phase-out date of 2027-01-01 (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes; market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API).

R410A is smaller but still material: 1,896 listings at GWP 1,924, the highest GWP among the main declared refrigerants in this dataset (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes; market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). That makes it a classic high-impact legacy exposure. By contrast, R290’s 537 listings at GWP 0 leave it commercially visible but still niche in declared volume (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes; market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API).

This is consistent with the current brand structure of the catalog, where large-volume suppliers such as Daikin Europe N.V. with 14,668 models, Mitsubishi Electric Europe B.V. with 5,575, JOHNSON CONTROLS HITACHI AIR CONDITIONING EUROPE SAS, SUCURSAL EN ESPAÑA with 5,207, Bosch Thermotechnik GmbH with 3,602 and Ariston SpA with 2,618 shape much of the searchable market universe (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). The corpus does not provide brand-by-refrigerant splits, so it cannot assign each manufacturer’s exposure numerically.

What buyers and installers should infer from the current market

Three practical conclusions follow from the data.

First, the average EU heat-pump market is not yet a natural-refrigerant market. Natural refrigerants hold 3.27% of listings, while R32 alone covers 22.85% (market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). Anyone assuming the transition is already complete is reading future intent into a still legacy-heavy registry.

Second, leakage-risk exposure in 2026 is concentrated where listing counts are largest, not where the best climate profile sits. R32 combines high declared volume, GWP 771 and a 2027 phase-out date, while R410A combines lower but still material volume with GWP 1,924 (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes; market_index_snapshot / Househeating Pulse · Market Index v1, computed from EPREL Public API). That does not make every R32 unit a poor choice, but it does mean installed-base risk remains anchored in HFC-heavy stock.

Third, efficiency leadership and refrigerant leadership are not fully visible in the same EPREL extract. The type data shows strong average performance for water-water and solid market-wide performance for air-water, while the top-model table shows SCOP up to 7.0, but the refrigerant field is missing for those leaders in this corpus (type_efficiency / EPREL Public API · type aggregation; top_models / EPREL Public API via Househeating Pulse catalog). Buyers comparing shortlist options should therefore combine the leaderboards, the methodology notes, and the model-level catalog filters rather than rely on refrigerant code alone.

For installers, the strongest current inference is narrow but useful: lower-GWP refrigerants such as R290 reduce climate exposure if leakage occurs, but A3 flammability means the installation regime matters just as much as the refrigerant label (refrigerant_universe / IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes). For buyers, the right question is less “which refrigerant wins?” than “which combination of type, efficiency, refrigerant and regulatory durability fits this building?” Our sizing calculator, payback calculator and climate-fit analyzer are better tools for that decision than broad refrigerant folklore.

Sources

• Househeating Pulse · Market Index v1, computed from EPREL Public API — snapshot 2026-06-22
• EPREL Public API · type aggregation — snapshot 2026-06-22
• EPREL Public API via Househeating Pulse catalog — snapshot 2026-06-22
• IPCC AR6 GWP table; EU Reg. 2024/573 phase-out schedule; EPREL declared codes — snapshot 2026-06-22

Continue reading

• How to compare heat pump SCOP properly — Why headline SCOP figures can mislead without type, test condition and temperature context.
• Heat-pump refrigerants explained — A practical reference on GWP, flammability classes and the current EU refrigerant transition.
• Air-to-water heat pumps: what to check before buying — A shortlist framework covering output, noise, efficiency and installer constraints.
• How to use EPREL data without getting fooled — What the registry captures well, where fields go missing, and how to read catalog data critically.