GLM 5.2 Q1_S beats Qwen 27B Q8 in hobby benchmark

View on Reddit

GLM 5.2 at Q1_S quantization outperforms Qwen 3.6 27B at Q8 in an amateur single-sample comparison. The test is a hobbyist benchmark with n=1 and should be taken as indicative rather than rigorous.