Total: 1
The observation of collectivity in small and large collision systems challenges our understanding of thermalization and particle production. EPOS4 models this via a dynamical core--corona separation, where high-density regions form a collectively expanding core while low-density regions hadronize via string fragmentation. Its microcanonical core hadronization improves the description of transverse momentum and multiplicity-dependent observables. We present EPOS4 predictions for pp, O-O and Pb-Pb collisions, with and without UrQMD, showing non-universal $\langle p_T\rangle$ scaling, significant hadronic-phase effects, and system-size-dependent $R_{AA}$ suppression. Charged-particle and transverse-energy densities show participant scaling; the transverse energy per charged particle is systematically larger in O--O than in Pb--Pb at comparable participant fraction, indicating a harder effective production in the lighter system. Identified-hadron spectra harden with event multiplicity with mass ordering and increasing core fractions. The mean transverse momentum exhibits a strong system dependence, with the steepest multiplicity evolution in pp, demonstrating that $\langle p_T\rangle$ does not follow universal multiplicity scaling. The $p/π$ ratio shows an enhanced intermediate-$p_T$ region; the suppression of the integrated $p/π$ at the highest Pb--Pb multiplicities is reproduced only with UrQMD, highlighting hadronic-phase effects. The nuclear modification factor shows sizeable suppression in Pb--Pb and substantial suppression in central O--O collisions. Blast-wave fits exhibit the anti-correlation between $T_{\rm kin}$ and $\langleβ_T\rangle$, with UrQMD shifting the parameters towards lower $T_{\rm kin}$ and higher $\langleβ_T\rangle$. These results provide a timely baseline for Run~3 measurements and for constraining the onset of medium-like effects across system size.