We present an analytic study for subdiffusive escape of overdamped particles out of a cusp-shaped parabolic potential well which are driven by thermal, fractional Gaussian noise with a 1/ω 1−α power spectrum. This long-standing challenge becomes mathematically tractable by use of a generalized Langevin dynamics via its corresponding non-Markovian, time-convolutionless master equation: We find that the escape is governed asymptotically by a power law whose exponent depends exponentially on the ratio of barrier height and temperature. This result is in distinct contrast to a description with a corresponding subdiffusive fractional Fokker-Planck approach; thus providing experimentalists an amenable testbed to differentiate between the two escape scenarios. The theme of anomalous sub-diffusion and rate kinetics continuous to flourish over the last years. This topic is driven by the availability of a wealth of intriguing experimental data, ranging from anomalous diffusion in amorphous materials, quantum dots, protein dynamics, actin networks, and biological cells [1,2,3,4,5,6,7,8,9,10,11]. Suitable theoretical descriptions derive from continuous time random walks (CTRW) [1,12], the CTRWbased fractional Fokker-Planck (FFP)-approach [13,14], or the generalized Langevin equation (GLE) [15,16]. The GLE-subdiffusion implies power-law-correlated thermal forces (or fractional Gaussian noise (fGn) [17]) possessing infinite memory with a 1/ω 1−α power spectrum [7,18]. Such random forces emerge when coupling the system to sub-Ohmic thermal baths with spectral densities J(ω) ∝ ω α , 0 < α < 1 [16].