This paper introduces the Koala PRF, which maps a variable-length sequence of 64-bit input blocks to a single 257-bit output block. Its design focuses on achieving low latency in its implementation in ASIC. To construct Koala, we instantiate the recently introduced Kirby construction with the Koala-P permutation and add an input encoding layer. The Koala-P permutation is obtained as the 8-fold iteration of a simple round function inspired by that of Subterranean. Based on careful preliminary cryptanalysis, we made a variant of the Subterranean permutation by reordering and modifying it in a way that does not introduce any implementation overhead and enhances the cryptographic resistance of the resulting PRF. Indeed, we demonstrate that Koala exhibits a high resistance against integral, cube, division property, and higher-order differential attacks. Additionally, we compare the hardware implementation of Koala with the smallest latency with state-of-the-art low-latency PRF Orthros and Gleeok and the block cipher Prince in the same ASIC synthesis setup. Our results show that Koala outperforms these primitives not only in terms of latency but also with respect to various other performance metrics.