TRi-structural ISOtropic (TRISO) fuel is a promising fuel type for advanced reactor designs because of its greater chemical and mechanical stability compared to traditional fuel rods. However, computational simulation of TRISO-fueled reactors is challenging because of the so-called double heterogeneity structure with inherent small fuel particles in random locations dispersed in moderated compacts, which are dispersed through a reactor core, and the overall large number of fuel particles, which is especially difficult for deterministic codes and simulations. Previously explored few-group methods for simplifying deterministic simulations by homogenizing the fuel particles within a compact induced system eigenvalue errors of up to 5000 pcm. Here, we present a method of generating group cross sections that preserves the criticality eigenvalue even with a full compact homogenization transformation. Rigorous agreement between continuous energy and multigroup results is still lacking, however, with multigroup criticality eigenvalues incurring an error of ~1000 pcm. This cross-section generation method directly collapses from pointwise ENDF cross-section data to a broad group structure using a very highly detailed (>100 000 energy points) neutron weighting spectrum in the nuclear data processing code NJOY. A typical spectrum for each TRISO fuel particle layer is generated using Monte Carlo (MCNP) tallies in a TRISO unit cell. The cross sections are collapsed using NJOY and formatted using our OJOYU postprocessor. All cross sections for each nuclide are calculated with NJOY in parallel using our Message Passing Interface–enabled ETEN code. Our fuel compact homogenization is accomplished by our HMIX code, which homogenizes using standard group-weighted forward fluxes and material volumes. Comparing the heterogeneous system eigenvalue with the homogenized system eigenvalue using these cross sections, an average error of approximately +100 pcm was seen, with a large decrease in computation time. The difference between continuous energy heterogeneous eigenvalue and multigroup eigenvalue was approximately +1000 pcm for four neutron groups. This cross-section generation method presents a path forward for homogenized TRISO modeling with deterministic or multigroup Monte Carlo methods.