Subterahertz collective spin-resonance modes and field-adaptive reservoir computing in the chiral helimagnet <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"> <mml:msub> <mml:mi>Cr</mml:mi> <mml:mrow> <mml:mn>1</mml:mn> <mml:

Monoaxial chiral helimagnets (CHMs) host rich helical spin textures, including chiral soliton lattices (CSLs) with tunable periods, arising from the delicate interplay between Dzyaloshinskii-Moriya interaction (DMI), ferromagnetic exchange coupling, uniaxial magnetic anisotropy, and Zeeman energy. However, existing CHMs exhibit spin resonance modes in the gigahertz frequency range, limiting their potential for high-speed signal processing. Here, with the combination of ferromagnetic resonance, electron spin resonance, and magneto-Raman techniques, we gain access to uncover subterahertz CSL phonon modes in a typical CHM ${\mathrm{Cr}}_{1/3}{\mathrm{Ta}\mathrm{S}}_{2}$. Near the critical field, we identify nontrivial CSL phonon modes reaching 0.15 THz, while a uniform ferromagnetic resonance mode emerges at 0.375 THz in the forced ferromagnetic phase under 9 T. The CSL phonon frequency in ${\mathrm{Cr}}_{1/3}{\mathrm{Ta}\mathrm{S}}_{2}$ is a factor of five or six higher than that of isostructural ${\mathrm{Cr}}_{1/3}{\mathrm{Nb}\mathrm{S}}_{2}$ due to larger spin-orbit coupling-induced DMI. Through micromagnetic simulations, we obtain the frequency spectrum and resolve the spatial distribution of amplitudes, phases, and precession trajectories of the CSL modes, providing deep insight into the characteristics of each resonance mode. Furthermore, we demonstrate that physical reservoir computing (RC), which exploits the nonlinear collective spin dynamics and field-controlled hysteresis of these nontrivial spin textures, achieves exceptional performance in time-series prediction tasks. Our findings not only elucidate the intricate dynamic properties of CSL phases but also pave the way for exploring the potential application of CHM materials for subterahertz signal processing and neuromorphic computing.