In the field of wireless communications, particularly within the Internet of Things (IoT) domain, efficient and reliable modulation techniques are paramount. Gaussian Frequency Shift Keying (GFSK) is a favored modulation scheme due to its good spectral efficiency and its resilience to impairment of radio frequency (RF) circuits. However, the challenge of achieving high performance in terms of Bit Error Rate (BER) and Packet Error Rate (PER) while maintaining low complexity in the detector design remains a significant issue. Addressing this challenge, this paper introduces a novel K-best based GFSK detector that not only approaches the optimal performance in BER and PER, but also manages to do so with reduced computational complexity. Further, another key feature of our technique is the generation of Log-Likelihood Ratios (LLRs), which are crucial for subsequent soft channel decoding stages. By providing LLR outputs, our detector facilitates a more refined decoding process, enhancing the overall error correction capabilities of the system.