<p>In the context of increasing Arctic warming, it is essential to predict changes in Arctic sea ice area (SIA) and the onset of ice-free conditions. The main sources of uncertainty in the projections are internal variability, large variability in the model structures and uncertainty in future emission scenarios. To reduce the uncertainty, we evaluated 27 CMIP6 models over the historical period and selected the nine best performing models using a percentile-based method. For the three considered scenarios (SSP1-2.6, SSP2-4.5 and SSP5-8.5), the respective total reductions in March SIA by the end of the century are 5%, 11% and 36%. For September, the results show SIA reductions of 65%, 92% and 100% for the same scenarios. The potential timing of sustained ice-free conditions in the Arctic for three climate change scenarios is as follows: SSP1-2.6—not expected; SSP2-4.5—from September in the early 2060s to August in the mid-2070s; SSP5-8.5—five months by 2100, starting from September in the late 2040s and increasing to August from the mid-2050s, to October from the mid-2060s, to July from the early 2080s and finally to November from the mid-2080s. The analysis is complemented by an assessment of differences between model projections of ice-free conditions. Among the selected models, the CanESM5 model shows the largest decrease in sea ice area under forcing. By the end of the century, the CanESM5 model predicts that ice-free conditions will last for three months (August–October) under the SSP1-2.6 scenario, five months (July–November) under the SSP2-4.5 scenario, and nine months under the SSP5-8.5 scenario. The GFDL-ESM4, NorESM2-LM, and BCC-CSM2-MR models describe the ice that is most resistant to warming. They expect no ice-free conditions by the end of the century under the SSP1-2.6 scenario, occasional occurrences of ice-free conditions in August under the SSP2-4.5 scenario, and 3–5 ice-free months under the SSP5-8.5 scenario.</p> Graphical Abstract <p>Global warming is significantly impacting Arctic sea ice, with observed declines from approximately 8 million km² to 4.5 million km² in September between 1979 and 2024. This study evaluates future projections of Arctic sea ice area (SIA) using 27 CMIP6 models under three Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, and SSP5-8.5) from 2015 to 2100, with particular focus on ice-free conditions (SIA &lt; 1 million km²). Model performance is assessed by comparing monthly sea ice concentration data against satellite observations (1979–2014) using statistical metrics such as correlation coefficients, root mean square deviation, standard deviation, and the climate prediction index. Models are ranked based on their skill in reproducing observed seasonal and interannual variability (for year and every month from June to November), as well as spatial biases and trend differences at each grid point. The nine most skilful models were selected to form a high-performance sub-ensemble. The results indicate that, under the SSP1-2.6 scenario, ice-free conditions are unlikely. The SSP2-4.5 scenario projects ice-free conditions in September by the early 2060s, and in August—by the mid-2070s. The high-emission SSP5-8.5 scenario projects more rapid and extensive ice loss: September could become ice-free as early as the late 2040s, followed by August (mid-2050s), July (early 2080s), October (mid-2060s), and November (mid-2080s). Analysis of changes in sea ice concentration shows that under the SSP1-2.6 scenario, SIA declines by approximately 65% in September and by 5% in March by 2100. The moderate scenario projects a 92% loss in September and an 11% reduction in March. Under the high-emission SSP5-8.5 scenario, September sea ice is expected to disappear entirely by 2100, with a 36% decrease in March. Using a high-performance model ensemble enhances confidence in these projections.</p> <p></p>