This study investigates the application of transfer learning with a transformer model for variable-length satellite image time series (SITS) to improve early crop classification in Slovenia. The primary objective is to produce reliable agricultural land class predictions in a timely manner, reduce required interventions, and limit in-field visits. The dataset consists of Sentinel-2 satellite imagery and reference crop labels from 2019, 2020, and 2021. The study evaluates the adaptability of the model through fine-tuning in a real-world scenario with limited up-to-date reference data. The base model trained on a different year achieved an average F1 score of 82.5% for the target year without having a reference from the target year. To increase accuracy, an average of 48,000 samples are required in the target year. Using transfer learning, the pre-trained models can be efficiently adapted to an unknown year, requiring less than 0.3% (1500) samples from the dataset. The results show that transfer learning can outperform the baseline in the context of early classification with only 9% of the data after 210 days in the year. The study also highlights the importance of class distribution in training data and the potential for further improvements in handling unbalanced datasets.This study investigates the application of transfer learning with a transformer model for variable-length satellite image time series (SITS) to improve early crop classification in Slovenia. The primary objective is to produce reliable agricultural land class predictions in a timely manner, reduce required interventions, and limit in-field visits. The dataset consists of Sentinel-2 satellite imagery and reference crop labels from 2019, 2020, and 2021. The study evaluates the adaptability of the model through fine-tuning in a real-world scenario with limited up-to-date reference data. The base model trained on a different year achieved an average F1 score of 82.5% for the target year without having a reference from the target year. To increase accuracy, an average of 48,000 samples are required in the target year. Using transfer learning, the pre-trained models can be efficiently adapted to an unknown year, requiring less than 0.3% (1500) samples from the dataset. The results show that transfer learning can outperform the baseline in the context of early classification with only 9% of the data after 210 days in the year. The study also highlights the importance of class distribution in training data and the potential for further improvements in handling unbalanced datasets.