Abstract The Karelian craton contains abundant ∼2.45 Ga mafic dykes that are compositionally, temporally and, in some cases, spatially related to PGE-mineralised mafic-ultramafic layered intrusions. The dykes can be sub-divided into four groups, namely siliceous high-magnesian basalts (SHMB), gabbronorites (GBNO), low-Ti tholeiites and Fe-rich tholeiites. In this study, we group the SHMB and GBNO dykes to one group as SHMB group, and the two tholeiitic group dykes as tholeiite group, based on their similar geochemical and mineralogical features. In the SHMB group dykes, plagioclase has initial 87Sr/86Sr ratios of 0.7028 to 0.7036 and initial bulk-rock εNd values vary from −2.5 to −1.0, indicating moderate degrees of contamination with Archaean basement. Tholeiitic dykes show a less-radiogenic Sr isotope composition with an average initial 87Sr/86Sr ratio of 0.7024 and higher initial εNd values ranging from +0.3 to +1.7. Thermodynamic and geochemical modelling suggests that the SHMB group dykes could have formed by crustal contamination of a komatiitic magma at deeper crust followed by fractional crystallisation at shallower depth, whereas the tholeiitic group mainly experienced fractional crystallisation with less crustal contamination. Alternatively, the SHMB dykes may have been derived from a SCLM mantle whereas tholeiitic dykes from a plume mantle, or the two types of dykes derived from different part of a mantle plume with different melting degrees, though these latter two models are not favoured in this study. Based on trace element and isotope characteristics, the SHMB dykes are suitable candidates for the parental magmas to some of the Finnish PGE-mineralised intrusions (e.g., Penikat and Portimo), whereas the tholeiitic dykes may represent the parental magma of the Tsipringa layered intrusion in Russia. Both the SHMB and tholeiitic dyke types are fertile with regard to PGE, with up to 10–20 ppb Pt and Pd and mantle-like Cu/Pd ratios in their least evolved members, suggesting that the magmas remained sulphide undersaturated during mantle melting and en route to the upper crust. This interpretation is consistent with the fact that most of the ∼2.45 Ga Fennoscandian layered intrusions contain PGE mineralisation. Sulphide melt saturation in the dykes and layered intrusions was mostly attained after their final emplacement, likely due to crystal fractionation.